Whether it’s from a sports injury, whiplash, or a bump to the head, many patients with mild concussion don’t even realize their minor injury can, if untreated, cause lifelong severe health issues. Even if a patient goes to the ER with their injury, it’s estimated that 50% – 90% of concussion cases go without a formal diagnosis, putting them at risk of dangerous complications such as brain bleeds and cognitive impairment.
A new research collaboration between the USC Viterbi School of Engineering and the USC Leonard Davis School of Gerontology has harnessed a powerful machine learning model to predict concussion status in patients.
The work, led by Benjamin Hacker (B.S. ’24), has now been published in the Journal of Neurotrauma, the premier flagship journal on brain injury. Hacker collaborated with Andrei Irimia, an associate professor of gerontology, biomedical engineering and neuroscience at the USC Leonard Davis School of Gerontology.
A concussion is a form of traumatic brain injury that can cause temporary changes to the brain’s function. Hacker said that current clinical practice for concussion diagnosis often relies on basic cognitive tests such as the Glasgow Coma Scale, a tool used to assess a patient’s level of consciousness, responsiveness and memory. Yet, many mild concussion patients never lose consciousness and may not present with the traditional cognitive symptoms that would make them easy to diagnose. Hacker said that this existing testing was not sensitive enough to detect many milder cases.
“We saw an opportunity to fit in that space between ‘not a concussion at all’ and a concussion that’s severe enough that it is consistently being detected,” said Hacker, who authored the paper as a USC Viterbi undergraduate and is now a master’s student in the Mork Family Department of Chemical Engineering and Materials Science.
“A clinician,” he added, “is not necessarily going to order imaging and request an MRI for someone who’s presenting without any symptoms. The idea is for this to be a secondary method that can aid the clinician when a patient is exhibiting certain symptoms, but they don’t have a firm concussion diagnosis based solely on cognitive tests.”
Early concussion diagnosis is critical for avoiding the long-term complications of traumatic brain injury, or TBI, Irimia said.
“Many concussions have negative effects called sequelae that only become apparent when it is too late for clinical interventions to improve their outcome,” Irimia explained. “The time period soon after injury is thus the most suitable for treatment to be initiated. Diagnosing a concussion early provides the best opportunity and time frame for treatment.”
Building on brain scan data
Hacker said that he and his collaborators built their model by harnessing MRI brain scan data from healthy control samples and people with concussions. The imaging that the classifier is based on is known as diffusion-weighted imaging, which measures how fluid travels through the brain on different connection paths.
“This data quantifies the directionality of diffusion between different regions in the brain. It tells us how strongly linked these different nodes are. We then used machine learning to develop a classifier,” Hacker said. “We trained this classifier on a discovery sample to teach it how the connectivity matrices of healthy people and injured people differ. Then, when we gave it independent testing samples, it was able to detect which of these subjects were concussed and which were healthy, based on the patterns in the brain connectivity matrix and on the strengths of certain neural pathways.”
Hacker and his team discovered that their classifier model worked incredibly well, showing 99% accuracy in both the training and testing samples.
“This is a much higher accuracy than we’ve ever seen with a method like this,” Hacker said. “I think it’s because nobody had previously devised our exact pipeline of using diffusion-weighted imaging, turning it into a connectivity matrix, and then leveraging machine learning in a tailored way to discover what pathways are most affected by head trauma. It is certainly novel in that we haven’t had an imaging-based classifier for concussion that has been accurate enough to rely upon until now.”
The classifier was built using Bayesian machine learning, which Hacker described as a probabilistic system that creates a classification based on whatever feature has the smallest probability of being incorrect or misclassified according to knowledge of prior conditions.
“It uses the training data to determine what patterns you would expect to see for a healthy person and what patterns you would expect to see for an injured person,” Hacker said.
Being the lead author of published research in an esteemed journal is a unique achievement for an undergraduate student. For Hacker, who is returning to USC Viterbi in the spring to complete his master’s in materials engineering, undertaking undergraduate research within the USC Leonard Davis School of Gerontology may seem like a surprising pathway.
Hacker was initially paired with the Irimia Lab through the Center for Undergraduate Research in Viterbi Engineering (CURVE) program. He soon found his chemical engineering background was a perfect and unique fit for this type of brain research. Hacker was well versed in chemical engineering theory around the way fluids move in various environments. This background knowledge translated well to the brain research he soon found himself specializing in, and a fascination with machine and deep learning helped propel his desire to better understand the neural connectome.
“I came up with this idea, with (Irimia’s) help, and was drawn to it because learning about diffusion — one way by which water and other fluids move — is very chemical engineering-based. It’s the core of how this study works, in the way that these brain scans were done — tracking the way that water diffuses through the brain,” Hacker said. “It was an opportunity for me to take a lot of what I had learned concerning fluid mechanics and numerical analysis and then apply it to something completely different from the applications that were presented in class.”
Potential for patients
The classifier that the research team created has the potential to form the basis of a concussion diagnosis platform that could be applied in clinical settings.
“We feel that this work definitely has the potential to disrupt the field in a positive way and be impactful. That’s the most exciting part for me. I can’t wait to see what this leads to,” Hacker said.
Applying the tool in a clinical setting could one day reveal components of brain circuitry that have been damaged by injury, which could help to formulate tailored strategies for clinical treatment based on a patient’s age at injury, Irimia said. Better treatment for brain injuries could have major long-term benefits, including lowering the risk for dementia.
“We are very interested in how concussions alter the risk for neurodegenerative diseases, including Alzheimer’s disease. There is evidence that TBI increases Alzheimer’s risk, but it is currently impossible to predict which concussion victims will progress to the disease,” he said. “We hope that our neuroimaging research will reveal risk factors that increase Alzheimer’s risk after concussion.”
Hacker’s collaboration with the Irimia lab is a great example of how interdisciplinary cooperation can lead to exciting advances and encourage translation of basic science research, Irimia added.
“Gerontology has grown into an interdisciplinary science that benefits greatly from the contributions of engineers. Many of the most promising trends in medical care involve the use of deep learning to improve the welfare of older adults,” he said. “Engineering can bridge the gap between basic science and clinical practice in addressing the unique needs of older adults and in personalizing medical strategies.”
Additional reporting by Beth Newcomb
Top image: The pathways in the brain highlighted are those most affected by concussion (image credit Benjamin Hacker et al.)
