EPViz: A flexible and lightweight visualizer to facilitate predictive modeling for multi-channel EEG

Area of Science:

• Computational Neuroscience
• Neuroimaging Analysis
• Machine Learning in Neuroscience

Background:

• Scalp electroencephalography (EEG) is a key noninvasive method for real-time neural activity study.
• Traditional EEG analysis focuses on group-level statistics, but machine learning drives a shift towards predictive modeling.
• There is a need for specialized tools to visualize and validate these complex predictive models.

Purpose of the Study:

• Introduce a novel open-source software, the EEG Prediction Visualizer (EPViz).
• To facilitate the development, validation, and reporting of predictive modeling outputs in EEG research.
• To bridge the gap between computational neuroscience and clinical applications through enhanced visualization.

Main Methods:

• Developed EPViz as a lightweight, standalone Python package.
• EPViz enables loading PyTorch deep learning models for EEG data.
• Allows overlaying channel-wise or subject-level temporal predictions onto original EEG time series.

Main Results:

• EPViz provides tools for viewing, manipulating, and visualizing EEG data and model predictions.
• Users can save high-resolution images of predictions for publications.
• Includes spectrum visualization, statistical computations, annotation editing, and EDF anonymization.

Conclusions:

• EPViz addresses a critical need for advanced EEG visualization in predictive modeling.
• The user-friendly interface and comprehensive features support collaboration between engineers and clinicians.
• Facilitates the sharing and interpretation of clinical EEG data for research and practice.