Cognizance detection during mental arithmetic task using statistical approach

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Clinical diagnosis relies heavily on physiological data, with neuronal activity analysis presenting challenges.
• Machine learning offers promising approaches for detecting defects in neuronal-assisted activity.

Purpose of the Study:

• To develop a machine learning model for classifying electroencephalogram (EEG) patterns into active and inactive segments.
• To utilize EEG signals from the frontal lobe during arithmetic tasks for intelligence detection.

Main Methods:

• Collected and segmented EEG data, extracting mean and standard deviation as features.
• Employed correlation and Fisher score for feature selection between Fp1 and F8 regions.
• Utilized R-studio and a Support Vector Machine (SVM) with a radial basis function kernel for classification.

Main Results:

• Identified Fp1 and F8 as vulnerable regions for arithmetic activity through correlation analysis.
• Achieved a sensitivity of 92.5% using the SVM classifier with selected features.
• Demonstrated the model's capability to classify intricate EEG patterns.

Conclusions:

• The developed SVM model effectively classifies EEG data, offering a sensitive method for detecting cognitive states.
• This approach has potential applications in diagnosing a wide range of clinical problems and advancing brain-computer interfaces.