A team of Georgia State scientists has received a two-year, $875,110 grant from the National Institute of Mental Health to further develop a tool to help psychiatrists treat mood disorders.
The researchers are based at the Center for Translational Research in Neuroimaging and Data Science (TReNDS), a tri-institutional effort supported by Georgia State University, Georgia Institute of Technology and Emory University, focused on making better use of complex brain imaging data.
The technology is based on research by Vince Calhoun, Distinguished University Professor of psychology and a Georgia Research Alliance Eminent Scholar.
He also holds appointments in Electrical and Computer Engineering at Georgia Tech and Neurology and Psychiatry at Emory University, and is the founding director of TReNDS. The grant was awarded to Advanced Biomedical Informatics Group, LLC, a start-up company led by Jeremy Bockholt, who has collaborated with Calhoun for more than 15 years.
Calhoun and his team aim to use data from functional magnetic resonance imaging (fMRI) to help psychiatrists predict how patients will respond to medication. Using machine learning, their algorithm would analyze a patient's fMRI scan and compare it to scans from thousands of other individuals.
Based on the clinical outcomes of those with similar brain activity, the tool could then predict how a patient would likely fare on one medication versus another. A psychiatrist could use this report, along with other information, to decide which medicine to prescribe.
There are no biologically based clinical tools to diagnose mental illness, and as a result distinguishing between mood disorders such as bipolar disorder and depression can be challenging.
(It takes an average of six to 10 years for bipolar disorder patients to receive a proper diagnosis.) Finding a mental health treatment that works for patients is often a process of trial and error that can take months or years.
"This tool could give clinicians an objective window into a patient's brain, helping them make more tailored treatment recommendations," said Bockholt. "Regardless of the diagnosis, is the patient's brain more similar to someone who responded better to mood stabilizers or to someone who responded better to antidepressants?"
Using the grant, the team will refine the algorithm by feeding it additional data, including scans from a more diverse set of patients and scans from various types of fMRI machines.
On the previous data set, our tool was over 90 percent accurate in predicting medication outcomes, so that shows us that we're on the right track. By training the model on more data, it should perform better on a wider variety of patients."
Eric Verner, Associate Director of innovation at TReNDS and Project Co-investigator, Georgia State University
The researchers will also complete interviews with psychiatrists to learn more about how and when such a tool could be used in a clinical setting. They hope to submit their technology to the U.S. Food & Drug Administration for approval as a medical device.
"We're focused on a patient population that is difficult to diagnose and treat using current methods," said Calhoun. "It can be hard to know what type of medication would be helpful for them, if medication is warranted. This could help inform those decisions and get patients on the right medication sooner."