BrainSymphony: A parameter-efficient multimodal foundation model for brain dynamics with limited data
Moein Khajehnejad, Forough Habibollahi, Devon Stoliker, Adeel Razi
arXiv Preprint Archive June 23, 2025 Peer reviewed via arXiv
Summary
A lightweight foundation model called BrainSymphony integrates fMRI time series and diffusion-derived structural connectivity using parallel spatial and temporal transformer streams, a Perceiver module for compact embeddings, and a novel signed graph transformer for anatomical connectivity. Despite requiring less data, it outperforms larger models on prediction, classification, and network discovery benchmarks. Attention maps from an independent psilocybin dataset reveal drug-induced reorganization of cortical hierarchies, demonstrating interpretable and clinically relevant results.
Study at a glance
| Design | theoretical or philosophical paper |
|---|---|
| Key finding | BrainSymphony, a compact multimodal model, outperforms larger models on neuroscience benchmarks and reveals drug-induced cortical hierarchy reorganization in an independent psilocybin dataset. |
Abstract
Foundation models are transforming neuroscience but are often prohibitively large, data-hungry, and difficult to deploy. Here, we introduce BrainSymphony, a lightweight and parameter-efficient foundation model with plug-and-play integration of fMRI time series and diffusion-derived structural connectivity, allowing unimodal or multimodal training and deployment without architectural changes while requiring substantially less data compared to the state-of-the-art. The model processes fMRI time series through parallel spatial and temporal transformer streams, distilled into compact embeddings by a Perceiver module, while a novel signed graph transformer encodes anatomical connectivity from diffusion MRI. These complementary representations are then combined through an adaptive fusion mechanism. Despite its compact design, BrainSymphony consistently outperforms larger models on benchmarks spanning prediction, classification, and unsupervised network discovery. Highlighting the model's generalizability and interpretability, attention maps reveal drug-induced context-dependent reorganization of cortical hierarchies in an independent psilocybin neuroimaging dataset. BrainSymphony delivers accessible, interpretable, and clinically meaningful results and demonstrates that architecturally informed, multimodal models can surpass much larger counterparts and advance applications of AI in neuroscience.