The USC Viterbi graduate’s research could provide important insights into quickly diagnosing traumatic brain injury with 99% accuracy.
CT images of the brain. Image Credit: Photo/Pexels
Whether it’s a sports injury, a blow to the head, or a blow to the head, many patients with minor injuries don’t even realize that if left untreated, their minor injury can last a lifetime. Can cause serious health problems. Even if a patient does go to the ER with their concussion, it is estimated that 50% – 90% of concussion cases go without a formal diagnosis, leaving them vulnerable to brain hemorrhage and cognitive impairment. There is a risk of dangerous complications like.
A new research collaboration between the USC Viterbi School of Engineering and USC Leonard Davis School of Gerontology A powerful machine-learning model has been used to predict a patient’s concussion status.
The work led by Benjamin Hacker (BS ’24) is now published Journal of NeurotraumaThe leading journal on brain injury.
Concussion is a form of traumatic brain injury that can cause temporary changes in brain function. Hacker said 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, reaction and memory. Nevertheless, many mild stroke patients never lose consciousness and may not present traditional cognitive symptoms making them easier to diagnose. Hacker said this existing test was not sensitive enough to detect many mild cases.
“We found a way to fit into that space between ‘no injury at all’ and an injury that is so severe that it is being continuously detected,” said Hanker, who wrote the paper as a USC Viterbi undergraduate and now a master’s student. Saw the opportunity.” in the Mork Family Department of Chemical Engineering and Materials Science.
“A physician,” he continued, “is not necessarily going to order imaging and request an MRI for a person with no symptoms. The idea is that it is a secondary method that can assist the physician when A patient may be exhibiting some symptoms, but they do not have a solid clinical diagnosis based on cognitive tests alone.
The hacker said he and his colleagues led andrei eremia, an associate professor of gerontology, biomedical engineering and neuroscience at the USC Leonard Davis School of Gerontology, created his model using MRI brain scan data from healthy control samples and people with concussions. The imaging on which the classifier is based is known as diffusion-weighted imaging, which measures how fluid travels through the brain on different connection paths.
“This data determines the direction of spread between different areas in the brain. This tells us how tightly connected 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 metrics of healthy people and injured people differ. Then, when we gave it independent test samples, based on patterns in the brain connectivity metrics and the strength of certain neural pathways, it was able to figure out which of these subjects had brain dysfunction and which were healthy.
Hacker and his team found that their classifier model worked incredibly well, showing 99% accuracy in both training and testing samples.
“This is far more accuracy than we’ve ever seen in a method like this before,” Hacker said. “I think it’s because no one had designed our exact pipeline in a tailored way to first use diffusion-weighted imaging, transform it into a connectivity matrix, and then leverage machine learning to figure out To determine which pathways are most affected by head trauma. “This is certainly new in that we have not had any imaging-based classifier for concussion that is accurate enough to rely on until now.”
The classifier was built using Bayesian machine learning, which Hacker describes as a probabilistic system that creates classifications based on the smallest probability of any feature being misclassified 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.
Becoming the lead author of a research published in a prestigious journal is a unique achievement for a graduate student. For Hacker, who is returning to USC Viterbi in the spring to complete her master’s degree in materials engineering, pursuing graduate research within the USC Leonard Davis School of Gerontology may seem like a surprising path.
Hacker was initially paired with the Iremia lab through the Center for Undergraduate Research in Viterbi Engineering (CURVE) program. He soon found that his chemical engineering background was perfectly suited and unique for this type of brain research. Hacker was well-versed in chemical engineering theory regarding the movement of fluids in various environments. This background knowledge translated well into brain research, he soon found himself an expert, and his fascination with machine and deep learning helped inspire his desire to better understand neural connectome.
“I came up with the idea with (Irimia’s) help, and was attracted to it because learning about diffusion – a way by which water and other fluids move – is very much based on chemical engineering. This is the core of how this study works, the way these brain scans were done – tracking the way water circulates through the brain,” Hacker said. “This was an opportunity for me to take a lot of what I had learned in relation to fluid mechanics and numerical analysis and then apply it to something completely different from the applications presented in the classroom.”
The classifier the research team created has the potential to form the basis of a concussion diagnosis platform that could be implemented in clinical settings.
“We think 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 happens with this,” Hacker said.
Source: USC
