EEG Disease Data Simulator and CNN-Attention Prototype

• A lightweight Python toolkit for:
  • generating synthetic EEG-like signals,
  • injecting disease-inspired EEG patterns,
  • extracting simple spectral features,
  • building a prototype labeled dataset,
  • supporting downstream deep learning experiments with a CNN + attention model.

---

Project Overview

• This project provides a modular prototype pipeline for synthetic EEG experimentation.
• It is designed for:
  • research prototyping,
  • educational demonstrations,
  • testing preprocessing and classification workflows,
  • generating small simulated datasets before moving to real EEG data.
• The codebase includes:
  • EEG signal generation,
  • disease pattern injection,
  • feature extraction,
  • dataset building,
  • a CNN-attention model definition.

---

What the Tool Currently Does

• Generates baseline synthetic EEG signals for a configurable set of EEG channels.
• Adds simple global artifacts/noise to the generated EEG.
• Injects simplified disease-inspired patterns for:
  • Alzheimer's disease (AD),
  • Parkinson's disease (PD),
  • Schizophrenia-like activity (SCZ).
• Extracts basic per-channel frequency-band features:
  • theta,
  • alpha,
  • beta.
• Builds a prototype dataset from generated signals.
• Defines a PyTorch CNN + attention model for EEG classification.

---

Important Notes / Corrections

• This is a prototype simulator, not a clinically validated EEG generator.
• The disease injection functions are currently simplified placeholders based on added sinusoidal activity.
• The current dataset_builder.py generates only AD-labeled samples.
• The current main_example.py demonstrates only:
  • EEG generation,
  • AD injection,
  • feature extraction.
• The CNN-attention model is defined, but the minimal modular files do not include a full standalone PyTorch training pipeline.
• The larger eeg_disease_data_simulator.py file appears to contain notebook-style experiments, visualization utilities, and extra exploratory code.

---

Project Structure

• eeg_config.py

  • stores default EEG configuration values,
  • defines the 10-20 channel system,
  • provides the time vector generator.

• eeg_signal_generation.py

  • generates baseline synthetic EEG signals,
  • creates global artifacts such as:
    • eye-blink-like low-frequency noise,
    • drift,
    • line noise.

• disease_injectors.py

  • injects simplified disease-inspired patterns into EEG:
    • inject_ad_pattern(...)
    • inject_pd_pattern(...)
    • inject_schizo_pattern(...)

• feature_extraction.py

  • computes simple frequency-band summary features:
    • theta,
    • alpha,
    • beta.

• dataset_builder.py

  • builds a small synthetic dataset from generated EEG samples,
  • currently labels generated samples as AD.

• model_cnn_attention.py

  • defines a 1D CNN + multi-head attention classifier in PyTorch,
  • outputs logits for 3 classes.

• main_example.py

  • demonstrates the basic pipeline from generation to feature extraction.

• eeg_disease_data_simulator.py

  • contains notebook-style exploratory code,
  • includes extended visualization and experimental sections,
  • should be treated as a research sandbox rather than the clean core API.

---

Installation

• Clone or copy the project files into one folder.
• Create a Python environment.
• Install the main dependencies:

pip install numpy torch matplotlib
  model = CNN_Attention(n_channels=19, seq_len=2560)

---