HP-GNN Implementation for Pediatric Seizure Prediction

Python source code implements a Hypergraph Neural Network (HP-GNN) model for seizure prediction. The package includes model architecture, training scripts, and evaluation tools for data from 22 pediatric patients. It was authored by Masoud Amiri and last updated in April 2026.

Use Cases

• Train the HP-GNN model on Kuramoto oscillator dynamics derived from EEG data for seizure prediction tasks.
• Evaluate model performance using the included metrics on the 22-patient pediatric cohort.
• Apply the data preprocessing scripts to convert raw EEG signals into 256-dimensional hypergraph structures.
• Use the provided PyTorch architecture to experiment with higher-order neural interactions via hypergraph convolutions.

Strengths

• Includes a complete, executable model implementation with training and evaluation pipelines.
• Model evaluation references a specific cohort of 22 pediatric patients.
• Code is released under a permissive CC-BY-4.0 license.

Limitations

• The dataset scope is limited, with code designed for a specific 22-patient cohort.
• The 19.7 KB size indicates a minimal implementation package without the underlying patient data.
• Performance details like the '84' accuracy metric lack full context and validation methodology.

Provenance

Source

Masoud Amiri via figshare.

Collection Method

Authored implementation code for a research model.

Freshness

Last updated April 2026.