A multimodal biomechanics dataset with synchronized kinematics and internal tissue motions during reaching.
Authors
Affiliations (10)
Affiliations (10)
- Department of Mechanical Engineering, MIT, Cambridge, MA, 02139, USA.
- Fraunhofer Portugal AICOS, Porto, 4200-135, Portugal.
- Comprehensive Health Research Center (CHRC), Porto, 4200-135, Portugal.
- Department of Electrical Engineering and Computer Science, MIT, Cambridge, MA, 02139, USA.
- Research Laboratory for Electronics, MIT, Cambridge, MA, 02139, USA.
- Department of Mechanical Engineering, MIT, Cambridge, MA, 02139, USA. [email protected].
- Institute for Medical Engineering and Science, MIT, Cambridge, MA, 02139, USA. [email protected].
- MIT.nano Immersion Lab, MIT, Cambridge, MA, 02139, USA. [email protected].
- Institute for Medical Engineering and Science, MIT, Cambridge, MA, 02139, USA. [email protected].
- MIT.nano Immersion Lab, MIT, Cambridge, MA, 02139, USA. [email protected].
Abstract
Tissue motions within body segments, such as the relative movements of muscles, fascia, and bone, remain largely unexplored despite their relevance to movement dysfunction, force transmission, and motor skill. Here, we present a time-synchronized multimodal dataset that bridges this gap by capturing both internal tissue dynamics and conventional biomechanical measurements during arm reaching. Thirty-six participants across three expertise levels (world-class athletes, regional athletes, and untrained individuals) performed slow, rhythmic reaching movements while we recorded data using B-mode ultrasound imaging, motion capture, electromyography, and accelerometry. The dataset includes processed signals, derived parameters (segmented reach events, tissue boundary motion, arm kinematics, tremor events, and muscle activation levels), and metadata. Notably, using the DUSTrack point-tracking workflow, we provide trajectories for 11 points across approximately 300,000 ultrasound frames from the upper arm. This resource enables at least three primary applications: (1) supervised training and benchmarking of deep learning models for point tracking in ultrasound videos, (2) development of ultrasound-based metrics for characterizing soft tissue mechanics, and (3) biomechanical investigation of how tissue-level dynamics support motor performance. All data, processing code, and tutorials are provided in accessible formats with documentation.