HP-GNN Explanation Agreement Metrics for Seizure Prediction

22 pediatric patient records form the basis for evaluating agreement between HP-GNN model explanations and clinical judgment. The dataset, created by Masoud Amiri, was last updated in April 2026 and is shared under a CC-BY-4.0 license. It contains metrics assessing the alignment of a hypergraph neural network's reasoning with expert clinical assessments.

Use Cases

• Evaluate the correlation between HP-GNN explanation features and clinical judgment scores for 22 pediatric patients.
• Benchmark new XAI methods against the provided agreement metrics for seizure prediction tasks.
• Analyze the impact of hypergraph convolution layers and Kuramoto oscillator dynamics on explanation fidelity in clinical contexts.

Strengths

• Provides specific evaluation metrics for 22 pediatric patient cases.
• Focuses on a specialized application of hypergraph neural networks (HP-GNN) in seizure prediction.

Limitations

• Very small dataset at only 5.5 KB, limiting statistical power and generalizability.
• Scope is narrow, focusing solely on agreement metrics for one specific model and clinical task.

Provenance

Source

Figshare, authored by Masoud Amiri.

Collection Method

Derived from evaluating an HP-GNN model's explanations against clinical judgment, likely from a research study.

Freshness

Last updated in April 2026.