Mood disorders, like many psychiatric conditions, have their onset in emerging adulthood and often last a lifetime. Mental illnesses have an excessively large loading of Disability Adjusted Life Years, making early intervention crucial for both individuals and society. Differentiating bipolar disorder (BD) from major depressive disorder (MDD) based on the Diagnostic and Statistical Manual (DSM) can be especially challenging if clear mania is absent. In fact, BD patients go an average of 6-10 years without the proper diagnosis, with 70% misdiagnosed with MDD instead. The challenge of identifying BD patients among depressed individuals is complex but critical because diagnosis determines treatment. The use of antidepressant (AD) medications in patients with BD can lead to worsening of illness. We have recently shown that, using resting fMRI, we can predict future medication class response with high accuracy (> 90%). In this project we will build on this work to generalize to a new MRI scanner and clinical assessment protocol. In addition we will develop, in consultation with multiple psychiatrists, a cloud-based tool to analyze and report the results from the brain imaging protocol and machine learning analysis, in a timely, meaningful, and interpretable manner. Results are expected to be an important step forward in the eventual development of clinical useful markers of mental illness. 2
Patients with mood disorders will often spend months to years on the wrong medication, which can also make them worse (e.g., putting individuals with bipolar disorder on antidepressants). There is a great need to develop biomarkers of treatment response in mental illness. This work will build on recent work from our group showing over 90% accuracy using resting fMRI predictors to further generalize the results to multiple scanners and develop an online portal to process and provide reporting of results (classification results) as well as processed data and citation information.
