Dissemination of a scalable and fair open-source motion analysis framework

The National Institutes of Health Office of the Director awarded Rehabilitation Institute of Chicago, doing business as Shirley Ryan AbilityLab, $161,754 on May 1, 2026 under the Trans-NIH Research Support program (CFDA 93.310) to develop and disseminate an open-source markerless motion capture (MMC) framework for biomechanical analysis and human pose estimation.

The recipient is transforming existing video acquisition and biomechanical analysis software—currently deployed in multiple NIH-funded research labs but requiring on-site engineering support and producing datasets with identifiable subject information—into a high-quality, broadly accessible open-source product. Work includes applying software engineering best practices to the codebase, simplifying implementation for clinical and research workflows, and establishing data de-identification processes to enable shareable outputs. The software performs whole-body tracking including arms and hands, addressing gaps in commercially available MMC systems that often fail to meet clinician and researcher needs and restrict customization due to proprietary constraints.

Performance occurs in Illinois and runs through April 30, 2029. The award is classified as a Project Grant, an assistance type supporting research by individual scientists or small teams.

Abstract

Advances in computer vision-based biomechanical analysis and human pose estimation have helped expand access to quantitative methods for movement analysis. These advancements have highlighted the utility of markerless motion capture (MMC) for overcoming the logistical challenges presented by marker-based systems. For example, MMC eliminates the cumbersome and time-consuming process of marker placement, and MMC systems can capture details that cannot be obtained by marker-based systems. However, commercially available MMC systems often fail to meet clinicians' and researchers' needs, and the proprietary nature of these systems prevents researchers from personalizing the systems and addressing the issues specific to their environment. Our team has developed MMC acquisition software designed to integrate into clinical workflows, as well as cutting-edge algorithms for biomechanical analysis, which are supporting several NIH-funded projects. Our collaborators using our software already report its many strengths, which speaks to the gaps we are filling, including whole-body tracking of arms and hands. While we have successfully deployed this software in multiple NIH-funded research labs, there are still barriers to making the system broadly accessible. For example, its implementation requires on-site engineering support. Additionally, output datasets and metadata are not readily shareable due to the fact that they contain identifiable information from subjects, such as raw videos, so additional processing steps are required for DE-identification of data. The overall objective of this research software engineer R50 proposal is to apply best software practices to this code and transform our current video acquisition and biomechanical analysis software, which works well for our research projects with experienced engineering support, into a high-quality open-source product that is easy to deploy in new labs and supports fair data sharing and reproducible RE-analysis of exported datasets. Specifically, we will ensure our software is truly accessible to a broad research and clinical audience by developing a robust, user-friendly, easy-to-deploy, open-source, and FAIR4RS markerless motion capture acquisition system and biomechanical analysis pipeline (aim 1). Further, we will facilitate researchers' ability to share and access movement data by implementing a data layout that produces DE-identified, fair- compliant datasets and metadata that can be easily imported and exported (aim 2). We anticipate that these improvements will democratize access to MMC by providing researchers and clinicians a robust and reliable software resource that can be easily implemented across a diverse range of settings, and this software will improve data sharing by producing datasets that can easily be imported to and exported from repositories for broad dissemination.