Psychologists Use Machine Learning to Help Identify Depression

The scientist is using the Stampede supercomputer at the Texas Advanced Computing Centre to train a machine learning algorithm that can recognize common issues among hundreds of patients employing Magnetic Resonance Imaging brain scans, genomics and other vital factors, to offer precise predictions of risk for those with anxiety and depression.

Scientists have long studied mental issues by analysing the relationship between brain structure and function in neuroimaging data. “One problem with that work is that it is mainly descriptive. The brain networks may appear to differ between two groups, but it does not inform us about patterns that actually identify which group you would fall into,” says Schnyer. “We are searching for diagnostic measures that are predictive for results such as vulnerability to dementia and depression.”

The machine that Schnyer and his group tested is known as Support Vector Machine Learning. The scientists offered a set of training examples each considered belonging to either healthy people for those who have been identified with depression. Schnyer and his group named the machine as labelled features in their data that were meaningful, and such examples were employed to train the system.

In the study, Schnyer analysed brain information from 52 treatment seeking participants with depression, and 45 healthy control participants. For a comparison, they matched a subset of depressed participants with individuals based on gender and age. “We feed in entire brain data or a subset and predict ailment classifications or any potential behavioural measure like measures of negative information bias,” he confirms. The research disclosed that DTI-derived fractional anisotropy maps can precisely classify vulnerable or depressed individuals versus healthy controls. It also revealed that predictive data is distributed across brain networks rather than being exceedingly localized.

The results obtained are promising, but still not clear enough to be employed as a clinical metric. But, Schnyer believes that by supplementing more data related to not just MRI scans but also form genomics and other classifiers, the system can perform much better.

“One of the primary benefits of this machine learning, compared to conventional approaches is that machine learning should enhance the likelihood that what we observe in the study will apply to novel and independent datasets. So, it must generalize to novel data,” says Beevers.

Schnyer and Beevers will expand their research to include data from numerous hundred volunteers from the Austin group who have been identified with depression, anxiety or a related situation Stampede 2 – TACC’s newest supercomputer which will come online later in 2017 and will be twice as powerful as the present system, will offer the enhanced computer processing power needed to include more data and achieve greater accuracy.