Enhancing Diagnostic Accuracy of Neurological Disorders Through Feature-Driven Multi-Class Classification with...

Area of Science:

• Neuroscience
• Medical Informatics
• Biomedical Engineering

Background:

• Neurological disorders (ND) present a global health challenge with overlapping symptoms complicating diagnosis.
• Current diagnostic methods for ND are often subjective, costly, and lack universal accessibility.
• Early detection of ND is critical for effective intervention and improved patient outcomes.

Purpose of the Study:

• To investigate the classification of multiple neurological disorders using electroencephalography (EEG) signals.
• To evaluate the efficacy of machine learning algorithms and feature selection for ND classification.
• To explore disease-to-disease classification as an alternative to traditional control-versus-patient frameworks.

Main Methods:

• EEG data from Alzheimer's, healthy controls, schizophrenia, MCI, and depression patients were analyzed.
• Feature extraction techniques were applied, followed by Least Absolute Shrinkage and Selection Operator (Lasso) for feature selection.
• Linear Discriminant Analysis (LDA) and other machine learning algorithms were used for two-class, three-class, and four-class classification tasks.

Main Results:

• Linear Discriminant Analysis (LDA) achieved 100% accuracy in distinguishing healthy controls from Alzheimer's disease.
• Multi-class classification (depression, MCI, schizophrenia) reached 84.67% accuracy.
• Frontal lobe EEG channels were frequently selected, indicating their importance in classifying neurological disorders.

Conclusions:

• EEG signal analysis combined with machine learning shows promise for classifying multiple neurological disorders.
• Disease-to-disease classification using EEG offers a novel approach for more precise clinical diagnostics.
• The findings support the development of more accessible and effective diagnostic tools for neurological conditions.