Functional Brain Network Identification in Opioid Use Disorder Using Machine Learning Analysis of Resting-State fMRI...

Area of Science:

• Neuroscience
• Radiology
• Machine Learning

Background:

• Opioid use disorder (OUD) neurobiology is crucial for developing effective treatments.
• Resting-state functional magnetic resonance imaging (rs-fMRI) offers insights into brain function.
• Traditional rs-fMRI analyses may not capture the full complexity of BOLD signals in OUD.

Purpose of the Study:

• To apply machine learning (ML) for time-frequency analysis of rs-fMRI BOLD signals in OUD.
• To differentiate individuals with OUD from healthy controls (HC) using neural activity patterns.
• To investigate the discriminative power of functional networks (DMN, SN, ECN) in OUD.

Main Methods:

• rs-fMRI BOLD signals were analyzed using time-frequency (wavelet) decomposition.
• Features were extracted from the default mode network (DMN), salience network (SN), and executive control network (ECN).
• 5-fold cross-validation classification (OUD vs. HC) was performed using ML, considering demographic factors.

Main Results:

• The DMN and SN demonstrated significant discriminative power between OUD and HC groups (p < 0.05).
• Mean F1 scores for DMN and SN were 0.7097 and 0.7018, respectively.
• Mean Area Under the Curve (AUC) values for DMN and SN were 0.8378 and 0.8755, respectively.
• Boruta ML analysis identified significant time-frequency detail coefficients across all three networks.

Conclusions:

• Time-frequency analysis of rs-fMRI BOLD signals, combined with ML, can effectively differentiate OUD subjects from HC.
• The DMN and SN are key functional networks with high discriminative potential in OUD.
• This approach offers a novel, data-driven method for understanding OUD neurobiology and may inform treatment strategies.