Discovering AI Applications for Traumatic Brain Injury Care (2024-2025)
The team examined how artificial intelligence (AI) can improve the management of traumatic brain injury (TBI), a leading cause of death and disability in the United States. Because TBI cases are highly variable and difficult to treat, the team sought to identify opportunities for AI to support clinical decision-making and patient care. To do so, students combined qualitative research, including interviews and shadowing across the Duke Health system, with quantitative analysis of electronic health records (EHRs) to inform the development of potential AI applications.
Through 28 clinician interviews across the emergency department, ICU, surgical units, and hospital floors, the team identified critical gaps in care that AI could help address. Providers highlighted opportunities for AI to triage patients in the ED, predict individualized prognoses earlier in the ICU, guide surgical interventions with better longitudinal data, and improve discharge education for families. These findings were compiled into an interactive application that maps the TBI care pathway. At the same time, students gained hands-on training in qualitative methods and clinical shadowing, deepening their understanding of how care is delivered at multiple points in the pathway.
On the quantitative side, the team worked with deidentified EHR data from Duke hospitals to prepare it for future AI modeling. Students cleaned and assessed datasets across demographics, labs, medications, and radiology reports, applying a quality assurance framework and validating usability with clinical champions. Their efforts yielded tools such as an algorithm for processing radiology text, visualizations of patient movement through care, and demographic analyses to detect bias. Together, the qualitative and quantitative tracks laid the groundwork for ethical, clinically relevant AI models that could enhance TBI care and potentially inform open-source research in the field.
Timing
Fall 2024 – Spring 2025
Team Outputs
Discovering AI Applications for TBI Care (Poster presentation at the Fortin Foundation Bass Connections Showcase, April 16, 2025)
Interactive application for TBI care pathways
Developed an organized codebase to carry out data cleaning
Algorithm for removing unwanted text from radiology reports
Sankey diagrams to visualize patient movement through care
Demographic analyses to detect bias
Image: Gene Activity After TBI, by Douglas Arneson and Drs. Gomez-Pinilla and Yang, UCLA, NIH Image Gallery, licensed under CC BY-NC 2.0