Effective Connectivity in Default Mode Network for Alcoholism Diagnosis

Area of Science:

• Neuroscience
• Medical Diagnostics
• Machine Learning

Background:

• Alcohol Use Disorder (AUD) is a chronic relapsing brain condition.
• Accurate AUD diagnosis is crucial for effective treatment and management.
• Current diagnostic methods rely on subjective criteria, leading to potential inaccuracies.

Purpose of the Study:

• To propose an objective assessment scheme for diagnosing AUD.
• To develop a machine learning model for classifying AUD based on brain activity.
• To overcome the limitations of subjective diagnostic criteria in AUD.

Main Methods:

• Electroencephalogram (EEG) data were collected from 31 AUD patients and 31 healthy controls (HC).
• Effective connectivity (EC) within the Default Mode Network (DMN) was calculated using partial directed coherence (PDC).
• A 3D Convolutional Neural Network (CNN) was employed for binary classification of AUD cases using EC matrices.

Main Results:

• The 3D-CNN model achieved 87.85 ± 4.64% accuracy in classifying AUD vs. HC using 5-fold cross-validation.
• Further validation with separate training and testing sets resulted in 100% correct classification of all testing subjects.
• The study demonstrates the potential of using brain connectivity patterns for objective AUD diagnosis.

Conclusions:

• The proposed EEG-based EC analysis combined with 3D-CNN offers a promising objective method for AUD diagnosis.
• This approach can potentially enhance the accuracy and reliability of AUD identification compared to conventional subjective methods.
• Objective diagnostic tools are vital for improving patient outcomes in Alcohol Use Disorder management.