Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic GNNs for Precise Seizure Detection and Classification from EEG Data

Arash Hajisafi, Haowen Lin, Yao-Yi Chiang, Cyrus Shahabi

TL;DR

NeuroGNN tackles the problem of precise seizure detection and subtype classification from EEG by constructing a dynamic NeuroGraph that fuses spatial, temporal, semantic, and taxonomic brain-contexts. It extends EEG analysis with six brain-region meta-nodes, BiGRU-based temporal encoding, MPNet-derived semantic embeddings, and a learned adjacency gate to integrate multiple similarity cues into the graph structure, followed by hierarchical pooling and an MLP classifier. A self-supervised pretraining objective enhances robustness under data scarcity, improving AUROC and weighted F1 on the TUSZ dataset, with ablations showing each context contributes to performance. The approach demonstrates strong practical impact for real-world EEG diagnosis by offering improved accuracy and resilience to limited labeled data, outperforming state-of-the-art GNN baselines on seizure detection and classification tasks.

Abstract

Diagnosing epilepsy requires accurate seizure detection and classification, but traditional manual EEG signal analysis is resource-intensive. Meanwhile, automated algorithms often overlook EEG's geometric and semantic properties critical for interpreting brain activity. This paper introduces NeuroGNN, a dynamic Graph Neural Network (GNN) framework that captures the dynamic interplay between the EEG electrode locations and the semantics of their corresponding brain regions. The specific brain region where an electrode is placed critically shapes the nature of captured EEG signals. Each brain region governs distinct cognitive functions, emotions, and sensory processing, influencing both the semantic and spatial relationships within the EEG data. Understanding and modeling these intricate brain relationships are essential for accurate and meaningful insights into brain activity. This is precisely where the proposed NeuroGNN framework excels by dynamically constructing a graph that encapsulates these evolving spatial, temporal, semantic, and taxonomic correlations to improve precision in seizure detection and classification. Our extensive experiments with real-world data demonstrate that NeuroGNN significantly outperforms existing state-of-the-art models.

Dynamic GNNs for Precise Seizure Detection and Classification from EEG Data

TL;DR

NeuroGNN tackles the problem of precise seizure detection and subtype classification from EEG by constructing a dynamic NeuroGraph that fuses spatial, temporal, semantic, and taxonomic brain-contexts. It extends EEG analysis with six brain-region meta-nodes, BiGRU-based temporal encoding, MPNet-derived semantic embeddings, and a learned adjacency gate to integrate multiple similarity cues into the graph structure, followed by hierarchical pooling and an MLP classifier. A self-supervised pretraining objective enhances robustness under data scarcity, improving AUROC and weighted F1 on the TUSZ dataset, with ablations showing each context contributes to performance. The approach demonstrates strong practical impact for real-world EEG diagnosis by offering improved accuracy and resilience to limited labeled data, outperforming state-of-the-art GNN baselines on seizure detection and classification tasks.

Abstract

Diagnosing epilepsy requires accurate seizure detection and classification, but traditional manual EEG signal analysis is resource-intensive. Meanwhile, automated algorithms often overlook EEG's geometric and semantic properties critical for interpreting brain activity. This paper introduces NeuroGNN, a dynamic Graph Neural Network (GNN) framework that captures the dynamic interplay between the EEG electrode locations and the semantics of their corresponding brain regions. The specific brain region where an electrode is placed critically shapes the nature of captured EEG signals. Each brain region governs distinct cognitive functions, emotions, and sensory processing, influencing both the semantic and spatial relationships within the EEG data. Understanding and modeling these intricate brain relationships are essential for accurate and meaningful insights into brain activity. This is precisely where the proposed NeuroGNN framework excels by dynamically constructing a graph that encapsulates these evolving spatial, temporal, semantic, and taxonomic correlations to improve precision in seizure detection and classification. Our extensive experiments with real-world data demonstrate that NeuroGNN significantly outperforms existing state-of-the-art models.
Paper Structure (21 sections, 3 figures, 4 tables)

This paper contains 21 sections, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology: NeuroGNN
  4. Node Features Generation
  5. Adjacency Matrix Generation
  6. Prediction using the Generated NeuroGraph
  7. Pretraining
  8. Experiments
  9. Experimental Setup
  10. Dataset and Data Splits
  11. Preprocessing
  12. Training
  13. Baselines
  14. Hyperparameter Setting
  15. Results
  16. ...and 6 more sections

Figures (3)

  • Figure 1: NeuroGNN Graph Construction
  • Figure 2: Seizure Detection and Classification using NeuroGraph
  • Figure 3: Visualization of graph embeddings for seizure classification test samples using trained Dist-DCRNN and NeuroGNN models. The colors represent the ground truth seizure labels.