270 | Predictive Brain Connectivity Analysis for Treatment Response in Major Depressive Disorder using Resting-State fMRI and Machine Learning

Theoretical and Computational Neuroscience

Author: DEBORA PATRICIA COPA | Email: dcopa@fi.uba.ar


1° Universidad de Buenos Aires, Facultad de Ingeniería, Buenos Aires, Argentina
2° Universidad de Buenos Aires, Facultad de Ciencias Exactas y Naturales, Departamento de Física, Ciudad Universitaria, Buenos Aires, Argentina

Major depression and other mood disorders often require pharmacological interventions, and the responses to these interventions can vary significantly among patients. This variability can lead to costly, prolonged treatments that can be burdensome for patients. Therefore, there is a desire to develop methods that can predict the effectiveness of medications before they are administered. In this study, a predictive modeling approach is proposed to assess the effectiveness of pharmacological treatments in patients with major depression. Machine learning algorithms were applied to resting-state functional magnetic resonance imaging (fMRI) brain images obtained under baseline conditions. The study highlights the utility of non-invasive measurements of brain activity in predicting treatment outcomes, indicating that specific functional connections support accurate prediction of treatment response. The method’s ability to forecast outcomes with different treatments in independent patient groups was evaluated. The results demonstrated that baseline brain connections, with feature preselection based on resting-state networks, exhibited a high capacity to predict treatment response. This suggests the feasibility of an automated system based on neurobiological data to enhance treatment decisions in psychiatry.
The need for future research to explore the widespread applicability of these algorithms and their utility in daily clinical practice is emphasized.