Machine Learning in Psychiatry

The COVID-19 pandemic has accelerated our integration of technology into our daily lives – whether for personal or professional use, there probably isn’t a day that goes by that you haven’t found a way to integrate it for improve your quality of life. I can’t imagine a day when I won’t use my phone to talk with others, my laptop to work, or an app to facilitate digital transactions. Technology has also made it possible to carry out remote appointments via the telehealth network, or via the remote administration of various tests and questionnaires.

However, there are places where we can accelerate the use of technology, along with machine learning, to improve diagnostic accuracy and treatment trajectory in psychiatry. Perhaps by avoiding the proven “trial and error” method, there could be an opportunity to shift our focus and invest in precision medicine.

Indeed, the neurobiology of psychiatric illnesses is not a “one size fits all” scenario, and in fact is heterogeneous, diverse, variable, and inherent in nature – simply because the brain, our behavior, and the external factors that influence these variables are dynamic. Due to these highly complex systems, which can be characterized as a composition of non-linear, multiplicative and cascading effects, there is a need to identify solutions that can handle these complex processes and produce accurate and active results.

Current drugs and therapeutics may only be effective for certain subgroups of individuals, thus referring to the “tried and true” method. This is not to invalidate currently available treatments, but rather to offer an alternative to improve the treatment decision process. In other words, there may be an opportunity to integrate deep learning to process these complex datasets.

If we can understand the complex neurobiological mechanisms that underlie various psychiatric disorders, then we can begin to identify shared diagnostic biomarkers and disease markers, which can help inform our understanding of different therapeutic response profiles. This information can form the basis of precision-based therapeutic interventions at the individual level. In order to achieve this goal, we would need large and powerful datasets, which could be obtained, for example, from electronic health record databases, social media platforms and momentary ecological assessments. , among other sources.

A few years ago, I conducted a narrative study on the use of momentary ecological appraisal to assess depressive symptoms. We found that passive smartphone-based apps can help assess depressive symptoms; however, at the time, this field was still nascent and underdeveloped. Now that time has passed and we are on a much faster technology growth phase, I think it would be prudent to revisit these tools (eg for diagnostic accuracy and treatment profiling).

For example, data that might be collected from a momentary ecological assessment might include physical activity, sleep quality, phone usage, social media usage, and GPS information. As such, these measures of physical and social activity can serve as a proxy for understanding the individual. However, integrating these types of data may come with some privacy, surveillance, and confidentiality concerns.

Taken together, there is tremendous potential to integrate machine learning into psychiatric care and also accelerate research and precision medicine in psychiatry.

Join Medscape’s new blogging initiative! We’re looking for doctors, nurses, physician assistants, specialists, and other healthcare professionals who want to share their expertise in one to two paid blog posts per month. Please email for more information.

Follow Medscape on Facebook, TwitterInstagram and YouTube

About Leanna Lui

Leanna MW Lui, HBSc, completed an HBSc Global Health Specialist degree at the University of Toronto, where she is now a Masters candidate. His interests include mood disorders, health economics, public health, and applications of artificial intelligence. In her spare time, she fences with the University of Toronto Varsity Fencing Team and the Canadian Fencing Federation.

Sherry J. Basler