Machine learning to tailor treatments in mental health (AI-MENT)

NTNU Health

Machine learning to tailor treatments in mental health (AI-MENT)

In psychiatry there is an urgent need for (i) reliable analytic tools to predict response/non-response to psychological treatment and (ii) to develop objective observational and diagnostic tools. Our ambitious aims are to (i) develop novel analyses and methods to predict response to psychological treatment using machine learning in order to avoid over-treatment and to offer personalized treatment, and (ii) improve observation and diagnosis in mental health wards, leading to better treatment and patient safety.

The Data and Artificial Intelligence (DART) group at IE and Trondheim Sleep and Chronobiology Research group (SACR) at IPH/St. Olavs Hospital cooperate in the studies.

Subjective sleep data from randomized controlled trials (2500 participants) are analyzed to find baseline predictors for response/non-response to treatment for an eight-week behavioral intervention to improve sleep in people suffering from sleep disturbances. The aim is to find participants in the studies that with a high probability will respond or not to the treatment, based on baseline information from each individual. By doing this, we will establish models that can be used in predicting response to psychological treatments for other disorders. One PhD, Stuart Gallina Ottersen, is working with these data together with the rest of the group. Stuart presented the project for the prime minister and the rector at the opening of the governments National Strategy for Digitalization June 6, 2023. 

The second set of aims are to improve and simplify diagnostics of sleep disorders and improve observation of and prediction of behavior in psychiatric hospital services using radar data with movement registration. We use objective radar data to observe patients in an acute psychiatric department (2500 nights). In addition, radar data is collected from smaller samples where the observations also include polysomnography and actigraphy. The datasets are of unprecedented size and quality, and are ideal for training machine learning algorithms and to develop new analysis methods. The project has a significant potential to change existing treatment and observation strategies in psychiatry. It may lead to easier, more cost-effective, and more precise diagnosing of sleep disorders, in addition to opening up a whole new field of using movement patterns as a tool to predict behavior. Post doc Hanne Siri Heglum and PhD Sophia Sylvester are working on these data together with the rest of the group. 

Read more about the research groups

Trondheim Sleep and Chronobiology Research group (SACR)

Data and Artificial Intelligence (DART) group:
Norwegian Research Center for AI Innovation
Norwegian Open AI Lab