Skip to content

Classification Of Alcoholic and Non-Alcoholic Subjects using Spiking Neural Network

Notifications You must be signed in to change notification settings

akash-55/EEG_classification

Repository files navigation

Classification of Alcoholic and Control Subjects using Spiking Neural Network

This project focuses on leveraging Spiking Neural Networks (SNNs) within the NeuCube computational framework to classify subjects with alcohol use disorders and control subjects based on EEG data.

Key Aspects:

1. Dataset

The EEG dataset consists of recordings from 122 control and alcoholic subjects, each undergoing 120 trials. Data were collected from 64 scalp-mounted electrodes.

2. Preprocessing

Using Python and bash scripts, the dataset was converted into NeuCube format and separated into two classes (Alcoholic and Control), with 100 samples from each class used for the experiment.

3. NeuCube Framework

NeuCube utilizes a spiking neural network (SNN) architecture designed to mimic the structural and functional aspects of specific brain regions. The SNN learns from spatio-temporal brain data (STBD), establishing connections between neuron clusters exhibiting sequences of neural activity.

4. Visualization

NeuCube's network analysis revealed notable differences in interconnectivity between control and alcoholic subjects. Visualizations demonstrated higher interconnectivity in the control group's brain activity.

5. Feature Extraction

Network analysis was used to identify and remove non-informative EEG channels, enhancing classification accuracy.

6. Results

Without feature extraction, the accuracy varied minimally between single and multiple training sessions. With feature extraction, accuracy significantly improved, achieving 80.20% after training the cube once and 75.25% after training it three times.

This project demonstrates the effectiveness of SNNs in handling complex neuroinformatic data, surpassing traditional machine learning methods in classifying alcohol use disorders. The use of NeuCube's feature extraction capabilities further enhanced the model's performance, showcasing the potential of advanced machine learning techniques in neuroinformatics.

About

Classification Of Alcoholic and Non-Alcoholic Subjects using Spiking Neural Network

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published