task-aware-brain-connectivity

What problem does it solve?

Researchers need robust methods to learn task-aware brain connectivity from fMRI time-series for downstream predictive tasks such as cognitive state classification or neurological disorder diagnosis.

Core Features & Use Cases

• DAG learning: Convert time-series into directed acyclic graph structures to capture causal brain interactions.
• Task-aware optimization: Jointly optimize connectivity learning with downstream prediction tasks using contrastive regularization.
• Graph neural encoder: Use a GNN to encode brain graphs and derive task-relevant embeddings for classification or regression.
• Use Case: Useful for fMRI analysis, brain network modeling, and diagnosis prediction in neuroscience research.

Quick Start

Provide your fMRI time-series data to TBDS and run a basic training pipeline to obtain task-aware connectivity embeddings.