Researchers use machine learning to help diagnose depression

Psychologists at the University of Texas used the Stampede supercomputer to identify patterns in neuroimaging data that may help predict depression.
By Amy Wallace  |  March 28, 2017 at 12:08 PM
share with facebook
share with twitter

March 28 (UPI) -- A new study from the University of Texas suggests machine learning with a supercomputer may help identify people susceptible to developing depression.

Depression is the leading cause of disability for people between the ages of 15 and 44, and affects more than 15 million American adults each year.

Researchers have studied mental illness by identifying the relationship between brain function and structure using neuroimaging data for years.

"One difficulty with that work is that it's primarily descriptive," David Schnyer, a cognitive neuroscientist at the University of Texas, said in a press release. "The brain networks may appear to differ between two groups, but it doesn't tell us about what patterns actually predit which group you will fall into. We're looking for diagnostic measures that are predictive for outcomes like vulnerability to depression or dementia."

Schnyer is using the Stampede supercomputer at the Texas Advanced Computing Center, or TACC, to train a machine learning algorithm that can identify similarities among hundreds of patients using magnetic resonance imaging, or MRI, genomics data and other factors to predict patients at risk for depression and anxiety.

Schyner and his colleagues used machine learning to classify people with major depressive disorder with nearly 75 percent accuracy.

Machine learning involves creating algorithms that can learn by building a model from sample data inputs and make predictions on new data.

The research team at the University of Texas used Support Vector Machine Learning to analyze data from 52 treatment-seeking participants with depression and 45 healthy participants.

Participants underwent diffusion tensor imaging MRI scans, which tag water molecules to determine the level to which those molecules are microscopically diffused in the brain over time. This diffusion measured in multiple spatial directions generates vectors for each voxel, 3D cubes that represent either structure or neural activity throughout the brain. The measurements are then translated into metrics that indicate the integrity of white matter pathways within the cerebral cortex.

Researchers compared the two groups and found statistically significant differences.

"We feed in whole brain data or a subset and predict disease classifications or any potential behavioral measure such as measures of negative information bias," Schyner said. "This is the wave of the future. We're seeing increasing numbers of articles and presentations at conference on the application of machine learning to solve difficult problems in neuroscience."

The study was published in Psychiatry Research.

Related UPI Stories
Trending Stories