Connectome-based predictive modeling of smoking severity using individualized structural covariance network in smokers.

Introduction: Abnormal interactions among distributed brain systems are implicated in the mechanisms of nicotine addiction. However, the relationship between the structural covariance network, a measure of brain connectivity, and smoking severity remains unclear. To fill this gap, this study aimed to investigate the relationship between structural covariance network and smoking severity in smokers. Methods: A total of 101 male smokers and 51 male non-smokers were recruited, and they underwent a T1-weighted anatomical image scan. First, an individualized structural covariance network was derived via a jackknife-bias estimation procedure for each participant. Then, a data-driven machine learning method called connectome-based predictive modeling (CPM) was conducted to infer smoking severity measured with Fagerström Test for Nicotine Dependence (FTND) scores using an individualized structural covariance network. The performance of CPM was evaluated using the leave-one-out cross-validation and a permutation testing. Results: As a result, CPM identified the smoking severity-related structural covariance network, as indicated by a significant correlation between predicted and actual FTND scores (r = 0.23, permutation Discussion: These results identified smoking severity-related structural covariance networks and provided a new insight into the neural underpinnings of smoking severity.