June 05, 2025
Traumatic brain injury (TBI) is a leading cause of disability and death globally. As the use of artificial intelligence (AI) technology in healthcare advances, new tools are emerging that have the potential to improve TBI diagnosis, treatment and outcomes. Since 1998, Mayo Clinic in Rochester, Minnesota, has served as a TBI Model System, one of 16 centers in the U.S. supported by the National Institute on Disability, Independent Living and Rehabilitation Research (NIDILRR). Currently led by Dmitry Esterov, D.O., M.S., Mayo Clinic's TBI Model System works to improve the lives of individuals who experience TBIs, their families and their communities. Dr. Esterov is a physiatrist and researcher and part of Mayo Clinic's internationally recognized brain rehabilitation program. One of Mayo Clinic's current TBI Model System research projects is exploring the use of an AI tool known as deep learning to analyze results of computerized tomography (CT) of the head in patients with TBIs.
Accurate prognostication of long-term functional outcomes early after TBI remains challenging. In severe TBI cases, uncertainty about prognosis can lead to withdrawal of life-sustaining treatment, despite evidence showing that some individuals can achieve meaningful recovery years after injury, regardless of the initial severity. For many individuals, understanding TBI prognosis early after injury is very important for rehabilitation planning and decision-making. Clinical predictors of outcomes after TBI, such as the length of post-traumatic amnesia, have been identified in the literature. However, currently there is no objective indicator of injury severity that can be obtained soon after injury to provide long-term prognostic information. This information is crucial for physicians, patients, caregivers and clinicians to discuss recovery, outcome trajectories and appropriate rehabilitation efforts.
CT of the head is the standard imaging technique for suspected acute TBI due to its sensitivity to life-threatening conditions, broad availability, low cost and rapid image acquisition. However, existing CT classification methods have only been found to correlate with acute outcomes such as death and need for advanced care, and they do not predict long-term clinical outcomes. Prior work within TBI Model Systems found that an objective indicator of injury severity based on traditional CT scan classification methods does not predict clinical outcomes after acute inpatient rehabilitation or at one year postinjury.
Deep learning (DL) models can identify neuroanatomic signatures in CT head data and could potentially be used to predict outcomes after TBI more accurately than existing methods. This study aims to use DL algorithms to extract disease signatures from head CT data to develop a prognostic biomarker for long-term functional outcomes in patients with TBIs.
The purpose of this study is to combine clinical data from participants in Mayo Clinic's Traumatic Brain Injury Model System data set with their digital CT scans to inform a DL model of anatomical injury. Mayo Clinic is partnering with three other TBIMS centers – Spaulding Rehabilitation /Mass General Brigham, Indiana University and University of Alabama.
This study explores what could be a significant advancement in the ability to predict long-term outcomes for patients with TBIs. By leveraging DL models and extensive clinical data, it aims to provide a reliable prognostic biomarker that can be used early in the treatment process. This will enable more-informed discussions with patients and their families about recovery trajectories and help determine the appropriate intensity of rehabilitative efforts, ultimately improving patient care and outcomes.
"This study involves partnering with Mayo Clinic's Digital Innovation Lab and other TBI Model System centers," explains Dr. Esterov. "We are attempting to unite clinical information from a projected 2,500 patients who experienced moderate to severe TBI with their digital CT data to inform a deep learning model of anatomical injury severity."
This research draws on the experience of practicing Mayo Clinic clinicians and technologists involved in development and implementation of deep learning-based clinical decision support systems. Mayo Clinic's TBI Model System researchers have led and collaborated on multiple groundbreaking clinical research efforts that have helped propel advances in epidemiology, practice management, treatment outcomes, behavioral interventions and remotely provided rehabilitation.
"This study involves partnering with Mayo Clinic's Digital Innovation Lab and other TBI Model System centers. We are attempting to unite clinical information from a projected 2,500 patients who experienced moderate to severe TBI with their digital CT data to inform a deep learning model of anatomical injury severity."
TBI Model System research focuses on the development, delivery and evaluation of innovative services that address patients' long-term needs, including care coordination and community reintegration. Mayo Clinic and other TBI Model System centers maintain a national database of individuals with moderate to severe TBIs and follow them for 30-plus years after injury.
For more information
Traumatic Brain Injury (TBI) Model System. Mayo Clinic.
Refer a patient to Mayo Clinic.