Building the Future of Youth Concussion Research Through Smarter Data

Concussions in youth athletes have long been a difficult problem to solve because the injury itself hides in plain sight. 

While some concussions are obvious, many are not — and countless smaller, “sub-concussive” hits may accumulate over years of play yet leave no obvious signs in the moment. The extent to which “sub-concussive” exposure across a practice or game, a season, or a career of play may meaningfully accumulate in ways relevant to the onset of an injury remains largely unclear.  Simply put, questions remain as to whether prior “sub-concussive” exposure may be creating an environment that leads to one individual experiencing an injury while another does not, even though they both experienced similar exposure events.  To understand what’s really happening inside the developing brain, researchers need long-term, multilayered data collected on the field, in clinics, and in everyday settings.

For more than a decade, Jason Luck, Ph.D., and his team in the Injury Biomechanics Laboratory (the Luck Lab) have been building exactly that kind of dataset in one of the longest-running studies of youth head impact exposure in the country. And now, thanks to pilot funding from the Duke Center for Computational and Digital Health Innovation, they’ve taken a major step forward: creating a modern, unified data pipeline that connects all the pieces of this massive, complex project.

The new pipeline is accelerating their ability to find, track, and eventually treat mild traumatic brain injury (mTBI) in young athletes.

“We were incredibly excited for this to be the first pilot funded through the Center,” says Amanda Randles, Ph.D., Director of the Duke Center for Computational and Digital Health Innovation. “This project is a perfect example of what we mean by Find, Track, and Treat, that investing in the data infrastructure needed to uncover subtle signals, follow them longitudinally, and ultimately translate them into better tools for clinical decision-making. Enabling this kind of connected multimodal analysis is exactly why the Center exists.”

A Long-Running Study Gets a Modern Data Pipeline

For 11 years, the Luck Lab has followed youth athletes ages 5 to 18 to understand how head impacts — ranging from severe concussions to seemingly harmless bumps — affect eye movement, cognition, and long-term neurological health. 

Participants complete an array of oculomotor tests (smooth pursuit, pro-saccades, anti-saccades, and memory-guided saccades), clinical concussion assessments, symptom surveys, athletic activity logs, and physical measurements of the head and neck. A subset of athletes at the high school level also wear the DASHR (Data Acquisition System for Head Response), an in-ear sensor that records detailed head-impact kinematics in real time.

The result is a rich, multimodal dataset covering entire seasons and, for many athletes, entire youth sports careers. The full value of that data, however, has been hard to access without a unified way to organize it.

“All of these data, due to their multimodal nature, have been stored in various formats and not easy to connect storage locations, making it difficult to draw connections across measures,” Dr. Luck says.

The Center’s pilot funding supported the development of a robust data warehouse and automated ingestion pipeline, which now allows the team to bring all these data streams into one place, structure them consistently, and analyze them as a unified whole.

The Stakes for Young Athletes

The consequences of missed brain injuries can be serious. 

“In children, brain injury is complex and common and is a leading source of disability and death,” Dr. Luck says.

Sports-related concussions account for 30 to 60% of all pediatric concussions, affecting up to 1.9 million children per year. Yet clinicians still lack reliable, objective tools to diagnose mTBI or quantify the effects of repeated sub-concussive impacts.

If researchers can identify measurable changes in eye movement, they may be able to detect brain injury earlier, more consistently, and across a wider spectrum of severity. The potential impact is enormous: safer return-to-play decisions, better long-term monitoring, and improved understanding of how the developing brain responds to repeated hits.

Thanks to the pilot funding, the team now has a fully functioning pipeline that:

• Uploads raw DASHR data and automatically detects impacts
• Parses wearable time-series data into standardized formats
• Digitizes hardcopy forms/surveys
• Processes eye-tracking data, transforming proprietary vendor outputs into formats suitable for analysis
• Allows rapid querying 

“In the past, we have largely relied on directly accessing and processing small subsets of the raw data across one or two modalities as needed, rather than working with multimodal or multiyear datasets.  The lack of a comprehensive pipeline with the built-in capability to access across our highly diverse dataset severely limited our ability to leverage the connectedness of the data to engage specific research questions,” Dr. Luck says. “Through this pilot study, we are now able to batch process data using our pipeline flows and can automate processing to trigger on newly uploaded datasets and ultimately engage highly multimodal and multiyear aspects of the data.” 

This shift unlocks new research possibilities by ensuring the team can analyze their dataset holistically, at scale, and with the consistency needed for advanced modeling.

A Framework With Room to Grow

Although designed for concussion research, the new data framework is broadly applicable. 

“We use the DASHR primarily in head injury research, but have laboratory versions of the system that can be placed in various locations around the body,” Dr. Luck says. “This pipeline would facilitate preprocessing and querying of these data as well.”

The pipeline is, in effect, a general-purpose data engine that happens to be demonstrated first in youth concussion science.

The team is now focused on expanding the pipeline’s capabilities. This effort is spearheaded by Mitchell Abrams, Ph.D., a postdoctoral researcher in the Luck Lab.  Upcoming enhancements include:

• Integrating eye-tracking data directly into the warehouse
• Automating digitization of hardcopy forms through optical character recognition and mark recognition
• Applying machine learning to improve impact detection and link multiple data modalities
• Developing analysis-ready datasets for collaborations and downstream modeling

As these tools evolve, the pipeline will continue to strengthen the connection between raw real-world data and the scientific insights required to keep young athletes safe.