An Interpretable and Predictive Connectivity-Based Neural Signature for Chronic Cannabis Use

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Cannabis use is increasing globally.
• While the psychiatric impact of cannabis use is known, reliable biomarkers for chronic use are lacking.
• This study addresses the need for objective measures of chronic cannabis consumption.

Purpose of the Study:

• To develop an interpretable decoding model for chronic cannabis use.
• To identify neural signatures of chronic cannabis consumption using functional magnetic resonance imaging (fMRI) and machine learning.
• To explore the relationship between brain network activity and cannabis craving.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used to collect brain activity data.
• Machine learning models were trained to classify chronic cannabis users (n=166) and nonusers (n=124).
• Network analysis identified key brain regions and networks predictive of cannabis use.

Main Results:

• The decoding model achieved high accuracy (∼80%) in differentiating chronic users from nonusers.
• Predictive brain regions involved motor, sensory, attention, and craving networks.
• Key brain networks strongly correlated with self-reported cannabis craving.

Conclusions:

• A novel, interpretable neural signature for chronic cannabis use was established.
• The findings demonstrate the potential of fMRI and machine learning for identifying cannabis use disorders.
• The identified neural patterns offer insights into the neurobiological underpinnings of cannabis craving.