BiomechAgent: An AI Agent for Biomechanical Analysis

Markerless motion capture is making quantitative movement analysis increasingly accessible. However, analyzing the resulting data remains a challenge for clinicians without programming expertise. BiomechAgent is a code-generating AI agent that enables biomechanical analysis through natural language. Users can query databases, generate visualizations, and interpret data without writing code.

Evaluating BiomechAgent's Capabilities

To evaluate BiomechAgent's capabilities, a systematic benchmark was developed covering data retrieval, visualization, activity classification, temporal segmentation, and clinical reasoning. BiomechAgent achieved robust accuracy on data retrieval and visualization tasks and demonstrated emerging clinical reasoning capabilities.

Biomechanically-informed instructions significantly improved performance over generic prompts. Integrating validated specialized tools for gait event detection substantially boosted accuracy on challenging spatiotemporal analysis where the base agent struggled.

Performance with Local vs. Cloud Models

BiomechAgent was tested using a local open-weight model instead of a cloud-based LLM. Performance was substantially diminished in most domains other than database retrieval. This highlights the trade-offs in using local computing resources versus more powerful cloud solutions, especially in scenarios requiring complex reasoning capabilities. For those evaluating on-premise deployments, there are trade-offs to consider (see /llm-onpremise).

In short, BiomechAgent makes the data from accessible motion capture much more useful and accessible to end users.