Application of the Random Forest Algorithm for Accurate Bipolar Disorder Classification.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Artificial Intelligence in Medicine

Background:

• Bipolar disorder (BD) diagnosis is challenging due to its complex nature and alternating mood states.
• Accurate and timely diagnosis of BD is crucial for effective treatment and patient outcomes.
• Electroencephalogram (EEG) signals offer a potential objective measure for psychiatric conditions.

Purpose of the Study:

• To investigate the efficacy of the Random Forest (RF) algorithm in classifying bipolar disorder patients from healthy controls using EEG data.
• To identify key EEG features indicative of bipolar disorder.
• To assess the potential of machine learning as a tool to support BD diagnosis.

Main Methods:

• Analysis of EEG data from 330 participants (euthymic BD patients and healthy controls).
• Extraction of EEG features including power in frequency bands, Hurst Exponent, and Higuchi's Fractal Dimension.
• Classification of participants using the Random Forest (RF) machine learning algorithm.

Main Results:

• The RF model achieved a high classification accuracy of 93.41% for BD detection.
• Recall and specificity of the model exceeded 93%, demonstrating robust performance.
• The model identified interpretable physiological markers associated with BD from EEG data.

Conclusions:

• Random Forest algorithm shows significant potential as a reliable and accessible tool for supporting bipolar disorder diagnosis.
• EEG-based machine learning classification can complement traditional diagnostic methods, potentially reducing delays and improving accuracy.
• This approach represents a step towards precision psychiatry, integrating AI for better mental health disorder management.