Cross-Modal Epileptic Signal Harmonization: Frequency Domain Mapping Quantization for Pre-training a Unified Neurophysiological Transformer

TL;DR

This paper introduces EpiNT, a Transformer-based pre-trained model that unifies EEG and iEEG analysis using frequency domain mapping quantization, improving classification tasks in epilepsy diagnosis.

Contribution

The paper presents a novel frequency domain mapping quantizer and a unified Transformer model for EEG and iEEG analysis, trained on extensive multi-modal neurophysiological data.

Findings

EpiNT outperforms baseline models on six classification tasks.

Pre-trained on over 2,700 hours of data from 1,199 patients.

Robust representation learning for epilepsy neurophysiology.

Abstract

Scalp electroencephalography (EEG) and intracranial EEG (iEEG) are vital for epilepsy diagnosis and treatment. Their unified analysis offers the potential to harness the complementary strengths of each modality but is challenging due to variations in recording montages, amplitude and signal-to-noise ratio (SNR), and frequency components. To address the aforementioned challenges, this paper introduces EpiNT, a novel Transformer-based pre-trained model for unified EEG and iEEG analysis. EpiNT employs channel-independent modeling with masked autoencoders (MAE) and vector quantization (VQ), along with a frequency domain mapping quantizer to capture crucial frequency features. Pre-trained on over 2,700 hours of multi-modal clinical neurophysiological data from 1,199 patients, EpiNT outperformed both randomly initialized models and other pre-trained methods on six downstream classification tasks, demonstrating robust representation learning capabilities. This work presents a promising approach for unified epilepsy neurophysiology analysis.