Psychedelics like psilocybin alter consciousness by reorganizing brain connectivity in a context-sensitive way. In the largest psychedelic neuroimaging dataset to date, 62 adults underwent functional MRI and EEG before and after ingesting 19 mg of psilocybin, during rest and naturalistic stimuli. Under psilocybin, brain signals during eyes-closed conditions became similar to those during eyes-open conditions, with increased global functional connectivity in associative regions and decreased connectivity in sensory areas. Machine learning linked subjective effects to structured neural activity patterns. Stronger self-dissolving effects were associated with more distinct neural representations and next-day mindset changes, revealing a state of 'embeddedness' where networks that usually segregate internal and external processing integrate coherently, aligning neural dynamics with context.
A lightweight foundation model called BrainSymphony integrates fMRI time series and diffusion-derived structural connectivity, enabling unimodal or multimodal training without architectural changes and requiring less data than larger models. It processes fMRI data through parallel spatial and temporal transformer streams, distills embeddings via a Perceiver module, and encodes anatomical connectivity with a signed graph transformer. The model outperforms larger counterparts on benchmarks for prediction, classification, and network discovery. Attention maps from an independent psilocybin dataset reveal drug-induced reorganization of cortical hierarchies, demonstrating interpretability and generalizability. The work shows that architecturally informed multimodal models can surpass much larger models, advancing AI applications in neuroscience.