Download dataset Contact

TAIR: Toolkit for Autonomous Interventions and Robotics

TAIR combines comprehensive open-access datasets and anatomical models with an open-source testbed. Use these resources to test, train, and evaluate autonomous robotic stroke intervention technologies for research and commercial applications.

Clinical data sets

Clinical data for stroke intervention research and development

The TAIR database includes vascular imaging data that represents the critical imaging modalities essential for neurovascular procedures. These modalities are used to visualize the vascular anatomy, identify pathology, and guide procedures. These targeted datasets support workflows connected with the TAIR testbed, and also support community-wide clinical and diagnostic research. A flexible permissive license allows use in academic, government, and commercial applications.

01 – CTA / MRA

CTA & MRA imaging

A large-scale multi-center compilation of Computed Tomography Angiography (CTA) and Magnetic Resonance Angiography (MRA). All data were collected as part of routine critical care. Prior to release, all data were rigorously de-identified through removal of PHI from the image metadata. Defacing algorithms were applied to remove potentially identifying facial structures.

Current release (11.3.2026)
500 studies
Final release (12.30.2027)
2,000 studies
Format
DICOM
License
Creative Commons Attribution 4.0 International (CC BY 4.0)

02 – Flouroscopy

X-Ray Flouroscopy Video

Videos of X-ray fluoroscopy procedures collected from a multi-center study during interventional procedures. Fluoroscopic videos are paired with CTA/MRA imaging from the CTA/MRA database for multi-modality research and development use cases. Prior to release, all data were rigorously de-identified through removal of PHI from the image metadata. Algorithms were applied to remove potentially identifying facial structures and burned in text.

Current release (11.3.2026)
150 datasets
Final release (12.30.2027)
1,250 datasets
Format
MP4
License
Creative Commons Attribution 4.0 International (CC BY 4.0)
Download the data
Coming in December 2026

03 – Anatomical Models

Segmented, centerlined, and annotated models, as well as statistical-shape models

This database includes 100 segmented, centerlined, and annotated CTA/MRA datasets, 50 annotated video fluoroscopy datasets, and XX models generated from population statistics and shape-modeling techniques.

Release
100 models + 1 population SSM Download models
Geometry
2,000 models + multi-parameter SSMs
Formats
VTK · NIfTI · OpenUSD · JSON graphs
Derivatives
Meshes, centerlines, SSM coefficients
License
Creative Commons Attribution 4.0 International License (CC BY 4.0)

TAIR simulator

Training and evaluation for stroke intervention simulations

Use the TAIR testbed to simulate stroke interventions across a range of patient anatomies and pathologies. The physics-based simulator represents soft-tissue mechanics, fluid dynamics, robot kinematics, and X-ray fluoroscopy.

First open-source release scheduled for March 2027

Access the open-source TAIR testbed and supporting resources.

Apache 2.0 license

Offline

Data preparation

  • Download and organize clinical imaging
  • Segment anatomy and prepare models
  • Select scenarios and intervention targets

Real-Time

Simulation

  • Configure anatomy, devices, and conditions
  • Run intervention and navigation scenarios
  • Capture repeatable trial data

Offline

Evaluation and analysis

  • Measure performance and outcomes
  • Review intervention efficiency and safety
  • Compare methods across scenarios

Built on proven open-source platforms

Rather than a bespoke one-off simulator, TAIR extends toolkits the team already develops and maintains — and contributes every improvement back upstream.

3D Slicer Anatomical model segmentation
TubeTK Centerline extraction
SlicerSALT Statistical shape modeling of population-level vasculature
Pulse Physiology Engine Patient physiology and hemodynamic modeling
iMSTK Vessel, clot, and device mechanics; interactive multiphysics
OpenUSD Scene composition — the patient-specific digital twin
VTK Registration, scene graph assembly, and visualization
DeepDRR Physics-based X-ray image formation and reconstruction
ROS2 Bidirectional Communication Interface
NVIDIA Isaac Sim Robot digital twin and HITL Testing

TAIR documentation

Everything documented, versioned, and open

Find technical guidance for using TAIR data and open-source software and review licensing and citation information.

Data

Dataset and anatomical model documentation

Each database includes a data manifest, including a list of files, data dictionary, and file sizes. The anatomic modeling software pipeline documentation can be found in the GitLab repository.

Platform

Testbed & API

The TAIR testbed documentation explains how the models are defined and implemented. It also includes class documentation, API definition and usage, and an SDK. Additional resources cover performance specifications, verification and validation cases and results, as well as version, license, and citation information.

Use cases

Verification, validation, and credibility

Working closely with the FDA and clinical collaborators, the testbed is undergoing rigorous verification, validation, and credibility assessments. The final report will be available at the conclusion of Phase I of the ARPA-H AIR program.

Coming early 2028

Working together

Clinicians, engineers, and scientists

A collaboration bringing technical and clinical expertise together across three areas of research.

Data

Clinical Lead

University of North Carolina at Chapel Hill

  • Kitware, Inc.
  • University of Pittsburgh Medical Center

Anatomic modeling

Technical Lead

Kitware, Inc.

  • University of North Carolina at Chapel Hill
  • University of Pittsburgh Medical Center

Testbed

Technical Lead

Kitware, Inc.

  • University of North Carolina at Chapel Hill
  • Florida State University
  • Johns Hopkins University
  • Lightside Surgical, Inc.
Kitware, Inc.
UNC Chapel Hill
University of Pittsburgh
FAMU–FSU Engineering
LightSide

Acknowledgements

This research was funded, in part, by the Advanced Research Projects Agency for Health (ARPA-H). The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the U.S. Government.

Questions?

Contact the TAIR team with questions about data, project resources, and testbed integration.

Contact the TAIR team

TAIRContact@kitware.com