Skip to content

yibeisita/neuro-ai-playground

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

64 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Neuro-AI Playground

A hands-on learning lab for exploring EEG, MEG, and fMRI data using Python, MNE, Nilearn, and Machine Learning.

This repository is designed for learners with a Computer Science / AI background who want to understand neuroimaging data analysis — from basic visualization to deep learning and multimodal integration.

What You'll Learn

  • How to load and visualize brain data (EEG, MEG, fMRI)
  • How to preprocess and clean neuroimaging signals
  • How to analyze brain activity (ERPs, GLMs, connectivity)
  • How to apply machine learning & deep learning to brain data
  • How to build reproducible and open-source pipelines

Each tutorial is standalone and focused on a single concept.

Neuroimaging Data Types

Modality What It Measures (Directly) Primary Use Case Resolution Strengths
EEG (Electroencephalography) Electrical Potential (Voltage) on the scalp from neural currents. Capturing the exact timing of brain events (e.g., event-related potentials). Excellent Temporal (milliseconds) / Poor Spatial
MEG (Magnetoencephalography) Magnetic Fields outside the head generated by neural currents. Localizing brain activity with high precision and timing (e.g., surgical planning). Excellent Temporal / Good Spatial
fMRI (functional Magnetic Resonance Imaging) Blood Oxygen Level Dependent (BOLD) signal, a measure of blood flow/metabolism. Mapping brain functions to specific anatomical structures (e.g., localizing language centers). Excellent Spatial (millimeters) / Poor Temporal

The core difference lies in the trade-off between when activity occurs (temporal resolution) and where it occurs (spatial resolution).

  • EEG & MEG are direct measures of neuronal electrical activity. They provide millisecond-level temporal resolution, allowing us to track the rapid sequence of cognitive processing.
    • MEG is superior to EEG in spatial resolution because magnetic fields pass through the skull undistorted, whereas the electrical signals of EEG are smeared and blurred by the skull.
  • fMRI is an indirect measure. It detects the resulting change in blood flow (the hemodynamic response), which is linked to neural activity but is slow (on the order of seconds). This slowness makes its temporal resolution poor.
    • However, fMRI excels at spatial resolution, providing detailed, millimeter-scale maps of activation across the entire brain structure.

Repository Structure

neuro-ai-playground/
│
├── README.md
├── requirements.txt
├── LICENSE
│
├── tutorials/            # Learning notebooks organized by skill
│   ├── 01_foundations/
│   ├── 02_preprocessing/
│   ├── 03_time_domain/
│   ├── 04_frequency_domain/
│   ├── 05_spatial_analysis/
│   ├── 06_connectivity/
│   ├── 07_machine_learning/
│   ├── 08_deep_learning/
│   └── 09_multimodal/
│
└── datasets/             # Data download scripts

Learning Roadmap

Level 1: Foundations

Learn to load and visualize neuroimaging data

Tutorial Focus Modality Key Skills
01.1 EEG Basics Load and plot EEG signals, compute PSD EEG File I/O, time series plotting, PSD analysis
01.2 MEG Basics Visualize raw MEG and evoked responses, topographies MEG Sensor layouts, evoked vs raw, topographic maps
01.3 fMRI Basics Display mean images, glass brain, deviation maps fMRI NIfTI handling, anatomical overlay, temporal deviations

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 2: Preprocessing

Clean and prepare data for analysis

Tutorial Focus Modality Key Skills
02.1 EEG/MEG Cleaning Filtering, ICA, artifact removal EEG/MEG Signal processing, ICA decomposition
02.2 fMRI Pipeline Motion correction, slice timing fMRI BIDS, preprocessing workflows

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 3: Time Domain Analysis

Analyze brain activity over time

Tutorial Focus Modality Key Skills
03.1 ERP Analysis Event-related potentials EEG/MEG Epoching, averaging, statistics
03.2 GLM Activation Task-based fMRI analysis fMRI General linear model, design matrices

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 4: Frequency Domain

Understand brain oscillations

Tutorial Focus Modality Key Skills
04.1 Time-Frequency Spectrograms, wavelets EEG/MEG Wavelet transforms, multitaper methods
04.2 Oscillatory Power Band-specific analysis EEG/MEG Alpha, beta, gamma rhythms

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 5: Spatial Analysis

Localize brain activity

Tutorial Focus Modality Key Skills
05.1 Source Localization Inverse solutions MEG/EEG Forward/inverse modeling, dipoles
05.2 ROI Analysis Region-of-interest extraction fMRI Atlases, parcellation, timecourses

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 6: Connectivity

Map brain networks

Tutorial Focus Modality Key Skills
06.1 EEG/MEG Connectivity Phase locking, coherence EEG/MEG Functional connectivity metrics
06.2 fMRI Networks Resting-state networks fMRI Correlation matrices, graph theory

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 7: Machine Learning

Decode brain states

Tutorial Focus Modality Key Skills
07.1 EEG Decoding Classification basics EEG Scikit-learn, cross-validation
07.2 fMRI MVPA Multi-voxel pattern analysis fMRI Searchlight, classification
07.3 Cross-Validation Proper validation strategies All Nested CV, permutation testing

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 8: Deep Learning

End-to-end neural networks

Tutorial Focus Modality Key Skills
08.1 EEG CNN Convolutional neural nets EEG PyTorch, CNN architectures
08.2 EEG RNN Recurrent neural nets EEG LSTM, temporal modeling
08.3 fMRI 3D CNN Volumetric deep learning fMRI 3D convolutions, attention

Status: ദ്ദി(ᵔᗜᵔ) Complete

Level 9: Multimodal Integration

Combine different brain imaging modalities

Tutorial Focus Modality Key Skills
09.1 EEG-fMRI Fusion Joint analysis EEG + fMRI Feature fusion, co-registration
09.2 MEG-fMRI Fusion Temporal + spatial integration MEG + fMRI Source space fusion

Status: ദ്ദി(ᵔᗜᵔ) Complete

Installation

1. Clone the repository

git clone https://github.com/YOUR_USERNAME/neuro-ai-playground.git
cd neuro-ai-playground

2. Create and activate a virtual environment (venv)

python3 -m venv .venv
source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Launch Jupyter

jupyter notebook

Then navigate to tutorials/ and open any notebook to get started!

Core Tools & Libraries

Purpose Libraries
EEG/MEG Analysis MNE-Python
fMRI Analysis Nilearn
Machine Learning scikit-learn
Deep Learning PyTorch
Multimodal Fusion EEG+fMRI feature fusion, co-registration workflows
Visualization matplotlib, seaborn
Data & Numerics nibabel, pandas, numpy, scipy

Example Datasets

All datasets can be downloaded through the notebooks in each corresponding tutorial.

Dataset Modality Description Size
MNE Sample MEG/EEG Auditory/Visual task (used in 7.1, 8.1, 8.2, 9.2) ~1.5 GB
Nilearn Development fMRI Resting-state fMRI (used in 6.2) ~10 MB
Haxby 2001 fMRI Visual object categories + VT mask (used in 7.2, 8.3, 9.1, 9.2) ~150 MB
OpenNeuro ds000117 fMRI Famous faces task ~50 GB
OpenNeuro ds003775 EEG Motor imagery ~5 GB

Contributing

Contributions are welcome! You can:

  • Add new tutorials
  • Improve existing notebooks
  • Add documentation or examples
  • Report bugs or suggest features

Please open a Pull Request or Issue on GitHub.

Resources & References

Official Documentation

Citation

If you use this repository in your research or education, please cite:

@software{neuro_ai_playground,
  author = {Wang Chen, Yibei},
  title = {Neuro-AI Playground: A Learning Lab for Neuroimaging Analysis},
  year = {2025},
  url = {https://github.com/YOUR_USERNAME/neuro-ai-playground}
}

License

This repository is licensed under the MIT License.
You are free to use, modify, and distribute for educational and research purposes.

Created with ♡ by Yibei Wang Chen

About

A hands-on learning lab for exploring EEG, MEG, and fMRI data using Python, MNE, Nilearn, and Machine Learning.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors