Binomial Logistic Regression and Artificial Neural Network Methods to Classify Opioid-Dependent Subjects and Control...

Area of Science:

• Medical data classification
• Neuroscience
• Machine learning in healthcare

Background:

• Logistic regression (LR) and artificial neural networks (ANNs) are established methods for medical data classification.
• Quantitative electroencephalography (QEEG) provides valuable data for understanding neurological differences in conditions like substance use disorders.

Purpose of the Study:

• To compare the efficacy of LR and ANNs in dichotomizing opioid-dependent patients from control subjects using QEEG absolute power values.
• To identify significant QEEG features for classification and evaluate the performance of both models.

Main Methods:

• QEEG absolute power values were calculated for delta, theta, alpha, and beta frequency bands in 75 opioid-dependent patients and 59 controls.
• LR was employed for direct interpretation, while ANNs utilized a genetic algorithm for feature selection.
• Classification performance was assessed using accuracy, area under the ROC curve, and Gini coefficient.

Main Results:

• Artificial neural networks (ANNs) demonstrated superior classification performance compared to logistic regression (LR).
• Both models achieved satisfactory results, particularly with absolute power measures in the beta frequency band.
• The study identified significant independent variables for classification in the LR model.

Conclusions:

• The developed methodology shows promise as a clinical tool for differentiating substance use disorder subjects.
• QEEG analysis combined with machine learning offers a valuable approach for medical data analysis and diagnosis.
• Further research can explore these methods for broader applications in clinical settings.