Salience Network Connectivity Predicts Response to Repetitive Transcranial Magnetic Stimulation in Smoking Cessation:...

Area of Science:

• Neuroimaging
• Machine Learning
• Addiction Science

Background:

• Tobacco use disorder (TUD) presents a significant public health challenge.
• Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive neuromodulation technique explored for smoking cessation.
• Integrating functional magnetic resonance imaging (fMRI) with machine learning (ML) offers potential for identifying therapeutic targets and evaluating rTMS efficacy in TUD neural networks.

Purpose of the Study:

• To investigate if large-scale network connectivity, assessed via fMRI, can predict the effectiveness of rTMS in promoting smoking cessation.
• To explore the predictive power of different large-scale brain networks (default mode, central executive, dorsal attention, salience, and reward networks) on rTMS treatment outcomes.

Main Methods:

• Acquisition of smoking cue-reactivity fMRI (T-fMRI) and resting-state fMRI (Rs-fMRI) data from 42 treatment-seeking smokers before and after 10 sessions of active or sham rTMS targeting the left dorsal lateral prefrontal cortex.
• Analysis of five large-scale brain networks' connectivity patterns before and after rTMS, and comparison between active and sham conditions.
• Application of neural network and regression analyses to correlate average network connectivity with rTMS treatment effectiveness.

Main Results:

• Regression analyses revealed that higher salience network (SN) connectivity in T-fMRI and lower reward network connectivity in Rs-fMRI predicted better smoking cessation outcomes following rTMS (p < 0.01, Bonferroni corrected).
• Neural network analyses identified the SN as the most significant predictor of rTMS effectiveness, with feature importance scores of 0.33 for T-fMRI and 0.37 for Rs-fMRI.
• These findings suggest that SN connectivity plays a crucial role in predicting rTMS treatment success for smoking cessation.

Conclusions:

• Both task-based and resting-state fMRI connectivity within the salience network (SN) can predict successful smoking cessation outcomes from rTMS, albeit in opposite directions.
• Machine learning models demonstrate potential for personalizing and targeting rTMS interventions for TUD.
• Replication in larger cohorts is recommended to validate these machine learning-derived findings due to the study's small sample size.