Speech biosignatures, subtle acoustic and linguistic markers are unique to each person and can change after injury and indicate neurological disruption.
AI Listens for Brain Injury
Speech biosignatures, subtle acoustic and linguistic markers are unique to each person and can change after injury and indicate neurological disruption.
11
Research from Florida International University aims to improve how concussions are detected by focusing on the human voice.
Concussions – a concern for athletes at every level from youth leagues to the Super
Bowl- are classified as mild traumatic brain injuries. Yet they can have serious consequences if undetected or improperly managed. It is estimated that more than 50 % of concussions in the United States go undiagnosed, and roughly 70 % occur in sports settings.
Sideline evaluations often rely on brief self-reported symptom checks, such as asking whether an athlete feels dizzy or has a headache. However, concussions do not appear on CT scans or other imaging, and diagnosis often depends on assessments of vision, eye movement, reflexes, and balance.
Although concussions often cannot be seen, Christian Poellabauer, a professor in the Knight Foundation School of Computing and Information Sciences at FIU, and his research team have shown that they can be heard. Poellabauer began studying the connection between speech patterns and traumatic brain injury more than a decade ago.
“ We were looking for ways of measuring or detecting concussions that are a little bit more foolproof,” he said.
His research examines speech biosignatures, subtle acoustic and linguistic markers unique to each person that can change after injury and indicate neurological disruption. Researchers collected voice samples from hundreds of high school and college athletes in boxing, tackle football, lacrosse, rugby, and cheerleading, comparing preseason and in-season data from athletes with confirmed concussions and those without.
Artificial intelligence identified differences in amplitude, frequency, and vocal vibration among individuals with brain trauma. The system now correlates voice changes with cases of brain injury with more than 90 % accuracy.
While some concussions present visible signs, such as the reflexive upper limb fencing response( an involuntary response where a person extends one arm and flexes the other), many are far more subtle.
“ It’ s very difficult to find those more subtle cases,” Poellabauer said.“ The concern is if you have too many concussions in a row, or if you keep playing and you get another one, that’ s going to have long-term effects on your health.”
Doctoral candidate Rahmina Rubaiat is working to identify a single word or sound for baseline and diagnostic testing. Simplifying the process would allow trainers to collect preseason samples and repeat the test following a potential injury.
Beyond sports, the research may also apply to neurological diseases and high-risk professions, offering a new approach to detecting brain injury through sound. g