Machine Learning to Infer and Control Brain State

Explore how machine learning techniques can identify and manipulate brain states for treating psychiatric disorders in this 29-minute webinar presented by Dr. David Carlson from Duke University. Learn about a novel statistical modeling approach that represents brain electrical activity as a superposition of latent electome networks—functional connectomes that collectively define predictive brain states associated with disease, behavior, or clinical outcomes. Discover how these interpretable networks, characterized by their spectral power and directional relationships between brain regions, enable the design of precise intervention protocols for targeted brain treatments. Examine case studies demonstrating the application of these methods to social aggression and anxiety, including the development of a closed-loop intervention protocol that selectively reduces aggressive behavior while preserving pro-social interactions in animal models—showing significant improvements over traditional open-loop stimulation approaches. Gain insights into ongoing research applying similar methodologies to anxiety disorders and explore continued methodological developments aimed at better designing and tracking intervention impacts. The presentation covers the intersection of probabilistic and deep learning approaches with neuroscience applications, emphasizing data-driven methods for generating testable hypotheses and designing confirmatory experiments in psychiatric treatment research.