written by Justin Park
DSI Studio is a free, open-source tractography analysis tool developed by Frank Yeh at the University of Pittsburgh. It is one of the most widely used desktop applications for diffusion MRI analysis, supporting fiber tracking, white matter atlas segmentation, connectome analysis, and data export. It runs on Windows, Mac, and Linux and requires no programming knowledge to operate. This evaluation covers DSI Studio in the context of its use as the 2D comparison condition in the Connected study, which compared AR spatial exploration of tractography data against DSI Studio's standard desktop interface.
Installation and Setup
DSI Studio is downloaded directly from https://dsi-studio.labsolver.org/ as a standalone application. No installation wizard is required on Mac — the app opens directly after download. The application is large (approximately 1–2GB) due to bundled atlas data and model weights. First launch requires no configuration. On Mac, Gatekeeper may block the app on first open, requiring a manual override through System Settings → Privacy & Security.
Ease of setup: straightforward for users comfortable with downloading and opening applications. No dependencies, no terminal required.
Interface and Navigation
DSI Studio uses a multi-panel desktop interface that is functional but visually dated. The main workflow is divided into four numbered steps visible on the home screen: NIFTI to SRC, Reconstruction, Fiber Tracking, and Tractography Atlas. For the purposes of this evaluation, the Fiber Tracking and Tractography Atlas steps are most relevant.
Navigation in the 3D viewport uses mouse drag to rotate, scroll wheel to zoom, and right-click drag to pan. The controls are responsive and consistent. Rotation is unconstrained, which allows viewing the brain from any angle but can lead to gimbal lock — a known limitation where certain rotation sequences produce disorienting flips in orientation. Two participants in the Connected study specifically mentioned this as a difficulty during the 2D condition.
The interface provides axial, sagittal, and coronal slice overlays that can be toggled on and off. These add anatomical context but require some familiarity with neuroimaging conventions to interpret correctly.
Tractography Atlas
DSI Studio's built-in Human Connectome Project atlas contains 318 named tract systems derived from population-averaged data. Tracts are loaded via Tracts → Load Built-In Atlas, which opens a scrollable list of all available bundles. Individual tracts can be checked or unchecked to isolate specific systems. The atlas is automatically segmented and labeled, making it accessible without requiring users to run their own fiber tracking.
Directional color coding is applied by default — red for left-right fibers, green for front-back, blue for up-down — matching the standard neuroscience convention. This is the same convention applied in the Connected AR build, ensuring visual comparability between the two conditions.
Tract names in the atlas follow a consistent naming convention (e.g. ProjectionBrainstem_CorticospinalTractL, Association_ArcuateFasciculusL) but are not always intuitive for users unfamiliar with neuroanatomy. Finding a specific tract by name in a list of 318 requires either prior knowledge of the naming convention or significant scrolling.
Data Export
DSI Studio exports tract data in several formats including .trk, .tck, and its proprietary .tt.gz format. The .tt.gz format is a compressed binary file using a custom TinyTrack encoding that is not readable by standard neuroimaging libraries such as nibabel. Decoding .tt.gz files requires either DSI Studio itself or a custom parser using scipy to read the underlying MATLAB v4 binary structure. This is a meaningful interoperability limitation for developers building downstream pipelines.
For the Connected study, a custom Python parser was written to decode .tt.gz files and convert streamline coordinates to normalized JSON format suitable for Unity and Three.js rendering.
Performance
DSI Studio handles large tractography datasets efficiently. The full 318-tract HCP atlas loads in under 30 seconds on a standard MacBook. Rendering thousands of streamlines simultaneously is smooth at interactive frame rates. Toggling individual tract systems on and off is near-instantaneous. Performance was not a limiting factor in any aspect of the Connected study.
Strengths
Free and cross-platform with no installation complexity. The bundled HCP atlas provides immediate access to labeled, population-averaged tract systems without requiring raw diffusion MRI data or processing pipelines. Directional color coding is applied by default and matches neuroscience standards. Mouse-based navigation is smooth and responsive for most viewing angles. The 2D flat interface is familiar to users of standard neuroimaging software and requires no hardware beyond a laptop.
Limitations
Gimbal lock during unconstrained 3D rotation can disorient users, particularly when trying to view the brain from specific angles. Tract names in the atlas are technical and not accessible to non-expert users without a reference guide. The .tt.gz export format is proprietary and not interoperable with standard neuroimaging libraries. The 2D flat display collapses depth information, requiring users to mentally reconstruct 3D spatial relationships from a projected view — a cognitive load that is particularly significant for tasks involving depth position or hemisphere laterality judgments. There is no built-in mechanism for collaborative viewing or shared annotation.
Comparison to AR Condition
In the Connected study, DSI Studio served as the 2D baseline condition against which Meta Quest 3 passthrough AR was evaluated. Participants using DSI Studio rated it higher on ease of navigation (3.75 vs. 3.50) and perceived understanding (3.88 vs. 3.50) compared to the AR condition, and it produced marginally higher overall accuracy (88% vs. 84%). However, participants who completed the AR condition first improved more when switching to DSI Studio (+12 percentage points) than participants who completed DSI Studio first and then switched to AR (+6 percentage points), suggesting that AR spatial exploration may build mental models that transfer to 2D tasks.
Qualitatively, participants described DSI Studio's navigation as smoother and more immediately intuitive, while noting that the flat view lost spatial context that the AR condition preserved. The most commonly cited limitation was that the 2D projection made depth and position judgments harder than the AR spatial view.
Recommended Use Cases
DSI Studio is well suited for researchers and clinicians who need to explore or analyze tractography data on a standard workstation. It is the appropriate tool for detailed quantitative analysis, connectivity matrix generation, and publication-quality tract visualizations. For educational contexts or spatial comprehension tasks with non-expert users, its 2D interface presents meaningful barriers that spatial AR/VR tools may address more effectively.
Version Evaluated
DSI Studio, Hou version (March 27 2026). Academic version. Evaluated on macOS with Apple Silicon.
References
Yeh, F.C. et al. (2022). Population-based tractography atlas. Nature Communications 13, 4933.
Yeh, F.C. et al. (2013). Deterministic diffusion fiber tracking improved by quantitative anisotropy. PLOS ONE 8(11).
DSI Studio documentation: dsistudio.labsolver.org