Using Quantitative Tools and Methodologies to Improve Mental Health

Using Quantitative Tools and Methodologies to Improve Mental Health

Mount Sinai’s new Center for Computational Psychiatry is among the first integrated centers in the world that investigates how quantitative tools and methodologies can improve mental health.

3 min read

The connection between the biology of the brain and mental health issues such as addiction and depression has been famously difficult to pin down. But combining algorithmic models with behavioral testing and neuroimaging may be about to change that. This emerging interdisciplinary field, known as computational psychiatry, may create a new framework for psychiatry in the years ahead, significantly enhancing diagnosis and treatment.

The Icahn School of Medicine at Mount Sinai’s new Center for Computational Psychiatry is among the first integrated centers in the world that investigates how quantitative tools and methodologies can improve mental health.

Xiaosi Gu, PhD, Director of the Center for Computational Psychiatry, and Associate Professor of Psychiatry, and Neuroscience at the Icahn School of Medicine, is not looking to replace clinicians or the one-on-one conversations they have with patients. Rather, she hopes to arm clinicians with algorithms and objective data to aid in their complex decision-making. Those decisions are typically informed by the Diagnostic and Statistical Manual of Mental Disorder and based on observed symptoms that are sorted by the physician into broad—and often arbitrary—mental health categories. 

  • Mount Sinai is pushing the boundaries in computational psychiatry.

Machine learning could transform that process by identifying biomarkers that objectively flag a wide range of mental health disorders. These algorithms rely on huge datasets of brain activities—datasets collected from EEG, MRI, and fMRI scans when patients are either at rest or performing a cognitive task. How effective could these biomarkers be? Most studies can get to a predictive accuracy level of 70 percent to 80 percent, but with larger datasets and better algorithms, this will likely improve.

“Despite advances in neuroimaging over the years, we still do not have reliable biomarkers for conditions such as depression or PTSD,” said Dr. Gu. “We now know more sophisticated computational methods can help us address this problem.”

  • Algorithms rely on huge datasets of brain activities.

Mount Sinai has actually been pushing the boundaries of this field for several years, and the new Center for Computational Psychiatry is expected to greatly accelerate that effort. One five-year study that just completed its third year, for example, is aimed at identifying through brain scans what kind of substance an individual is craving, be it nicotine, alcohol, marijuana, or another. “We simply feed patients’ fMRI data into a machine learning algorithm and, amazingly, it has been able to predict the type of dependency with around 80 percent accuracy,” said Dr. Gu. A second study underway at the Center is combining theory-driven models, fMRI, and social interaction games to better understand the neural and cognitive mechanisms that underlie autism spectrum disorder.

These two early projects—the first data-driven, the second theory-driven—underscore the dual scientific objectives of the new Center’s research. “Machine learning can give us valuable predictions about what disorder a patient has, and what medication is likely to work best,” said Dr. Gu. “But that is not enough. We need to know why it works if we are going to design better treatments in the future. And that is where theory-driven models can play a critical role. We need both approaches for computational psychiatry to make a real difference and deliver tangible results.”

Featured

Xiaosi Gu, PhD

Xiaosi Gu, PhD

Director of the Center for Computational Psychiatry, and Associate Professor of Psychiatry, and Neuroscience, at the Icahn School of Medicine