Ant Colony Optimization Based Feature Selection Method for QEEG Data Classification.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Machine Learning in Healthcare

Background:

• Biomedical signal analysis often necessitates feature selection for efficient data representation.
• Identifying relevant features is crucial for accurate classification of complex conditions like Major Depressive Disorder (MDD).

Purpose of the Study:

• To evaluate the efficacy of an Ant Colony Optimization (ACO) algorithm for feature selection in electroencephalogram (EEG) data.
• To enhance the classification performance of a Back Propagation Neural Network (BPNN) for MDD detection using selected EEG features.

Main Methods:

• ACO algorithm was employed for feature selection on 6-channel pre-treatment EEG data (theta and delta bands).
• The selected features were used to train a BPNN classifier for distinguishing MDD subjects (n=147).

Main Results:

• The BPNN achieved 91.83% overall accuracy and 95.55% sensitivity in classifying MDD subjects.
• Feature selection using ACO increased the Area Under the ROC Curve (AUC) from 0.8531 to 0.911.
• The algorithm reduced the feature subset from 12 to 5 features (Fp1, Fp2, F7, F8, F3 in the theta band).

Conclusions:

• ACO-based feature selection demonstrably improves BPNN classification accuracy for MDD.
• The study highlights the potential of combining ACO with BPNN for robust biomedical signal analysis.
• Further research comparing various feature selection algorithms and classifiers is recommended to validate this approach.