Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

STEAM-EEG: Spatiotemporal EEG Analysis with Markov Transfer Fields and Attentive CNNs

Jiahao Qin, Feng Liu

TL;DR

STEAM-EEG tackles the non-stationary, noisy nature of EEG signals by integrating Singular Spectrum Analysis (SSA) based trend-seasonal decomposition, parallel 1D-CNNs with a modified cross-channel attention, and Markov Transfer Field (MTF) imaging to produce informative visual representations. The framework transforms EEG time series into SSA components, learns cross-channel features with attention, and renders spatiotemporal patterns as MTF images processed by a 2D-ResNet with split-channel attention for classification. Across diverse EEG datasets, STEAM-EEG achieves higher accuracy and F1-scores than strong baselines, with ablation showing MTF imaging as a key contributor to performance and interpretability. These results suggest STEAM-EEG can enhance EEG pattern recognition and visualization, with potential extensions to multi-channel and real-time biomedical signal analysis.

Abstract

Electroencephalogram (EEG) signals play a pivotal role in biomedical research and clinical applications, including epilepsy diagnosis, sleep disorder analysis, and brain-computer interfaces. However, the effective analysis and interpretation of these complex signals often present significant challenges. This paper presents a novel approach that integrates computer graphics techniques with biological signal pattern recognition, specifically using Markov Transfer Fields (MTFs) for EEG time series imaging. The proposed framework (STEAM-EEG) employs the capabilities of MTFs to capture the spatiotemporal dynamics of EEG signals, transforming them into visually informative images. These images are then rendered, visualised, and modelled using state-of-the-art computer graphics techniques, thereby facilitating enhanced data exploration, pattern recognition, and decision-making. The code could be accessed from GitHub.

STEAM-EEG: Spatiotemporal EEG Analysis with Markov Transfer Fields and Attentive CNNs

TL;DR

STEAM-EEG tackles the non-stationary, noisy nature of EEG signals by integrating Singular Spectrum Analysis (SSA) based trend-seasonal decomposition, parallel 1D-CNNs with a modified cross-channel attention, and Markov Transfer Field (MTF) imaging to produce informative visual representations. The framework transforms EEG time series into SSA components, learns cross-channel features with attention, and renders spatiotemporal patterns as MTF images processed by a 2D-ResNet with split-channel attention for classification. Across diverse EEG datasets, STEAM-EEG achieves higher accuracy and F1-scores than strong baselines, with ablation showing MTF imaging as a key contributor to performance and interpretability. These results suggest STEAM-EEG can enhance EEG pattern recognition and visualization, with potential extensions to multi-channel and real-time biomedical signal analysis.

Abstract

Electroencephalogram (EEG) signals play a pivotal role in biomedical research and clinical applications, including epilepsy diagnosis, sleep disorder analysis, and brain-computer interfaces. However, the effective analysis and interpretation of these complex signals often present significant challenges. This paper presents a novel approach that integrates computer graphics techniques with biological signal pattern recognition, specifically using Markov Transfer Fields (MTFs) for EEG time series imaging. The proposed framework (STEAM-EEG) employs the capabilities of MTFs to capture the spatiotemporal dynamics of EEG signals, transforming them into visually informative images. These images are then rendered, visualised, and modelled using state-of-the-art computer graphics techniques, thereby facilitating enhanced data exploration, pattern recognition, and decision-making. The code could be accessed from GitHub.
Paper Structure (16 sections, 9 equations, 5 figures, 2 tables)

This paper contains 16 sections, 9 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Trend-Seasonal Decomposition using Singular Spectrum Analysis
  5. Parallel 1D Convolutional Neural Networks with Modified Attention Mechanisms
  6. Markov Transfer Field Imaging
  7. Two-dimensional Residual Convolutional Neural Network with Cross-Channel Split Attention
  8. Results
  9. Datasets
  10. Baselines
  11. Evaluation Metrics
  12. Classification Performance
  13. Ablation Study
  14. Visualization
  15. Discussion and Limitations
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Performance comparison of our proposed SOTA method against baseline approaches across various EEG datasets. The radar chart illustrates the superior accuracy of our method (in orange) compared to FCNN, EEGNet, DeepConvNet, and STFT-CNN across ten different EEG datasets.
  • Figure 2: The overall architecture of the proposed approach for enhanced EEG signal analysis.
  • Figure 3: Architecture of the two-dimensional residual convolutional neural network with cross-channel split attention for image feature extraction.
  • Figure 4: Sample of Trend-Seasonal decomposition with Singular Spectrum Analysis.
  • Figure 5: Feature extraction process: original MTF images and corresponding feature maps from the penultimate and final image feature extraction layers for both classes.