gnn-transformer-fusion

What problem does it solve?

This skill enables a principled approach to fuse non-Euclidean brain connectivity data with Euclidean clinical and demographic data to enable time-aware longitudinal predictions.

Core Features & Use Cases

• Multi-modal fusion: jointly encoding brain graphs and tabular data with cross-modal attention for integrated representations.
• Time-aware modeling: captures temporal dynamics across multiple visits using a Transformer-based fusion layer.
• End-to-end workflow: trainable, end-to-end pipeline for prediction tasks like disease progression, treatment response, and cognitive trajectory forecasting.

Quick Start

Load longitudinal brain connectivity and tabular data and run the GNN-TF pipeline to generate time-aware predictions.