Classifying and characterizing nicotine use disorder with high accuracy using machine learning and resting-state fMRI

Area of Science:

• Neuroscience
• Addiction Research
• Computational Psychiatry

Background:

• Cigarette smoking is a major cause of preventable death, with modest smoking cessation rates.
• Accurate methods to classify nicotine use disorder (NUD) neural features are needed for treatment optimization.
• Resting-state functional connectivity (rsFC) offers potential biomarkers for NUD.

Purpose of the Study:

• To apply machine learning to rsFC data for classifying NUD.
• To identify specific brain networks involved in NUD.
• To visualize the heterogeneity of NUD using rsFC.

Main Methods:

• Support vector machine classification applied to rsFC data from NUD patients (n=108) and controls (n=108).
• Multi-dimensional scaling used to visualize NUD heterogeneity.
• Analysis focused on five key resting-state networks.

Main Results:

• Machine learning models achieved 88.1% classification accuracy and an AUC of 0.93.
• Individuals with NUD showed weaker functional connectivity within identified networks compared to controls.
• NUD patients exhibited greater heterogeneity in rsFC compared to controls.

Conclusions:

• Machine learning classification of rsFC data is a valuable tool for understanding NUD neurobiology.
• This approach improves classification accuracy and reveals network-level heterogeneity in NUD.
• Findings support the development of data-driven, personalized smoking cessation strategies.