A deep learning model for characterizing altered gyro-sulcal functional connectivity in abstinent males with methamphetamine use disorder and associated emotional symptoms

Failure to manage emotional withdrawal symptoms can exacerbate relapse to methamphetamine use. Understanding the neuro-mechanisms underlying methamphetamine overuse and the associated emotional withdrawal symptoms is crucial for developing effective clinical strategies. This study aimed to investigate the distinct functional contributions of fine-scale gyro-sulcal signaling in the psychopathology of patients with methamphetamine use disorder and its associations with emotional symptoms. We recruited 48 male abstinent methamphetamine use disorders and 48 age- and gender-matched healthy controls, obtaining their resting-state functional magnetic resonance imaging data along with scores on anxiety and depressive symptoms. The proposed deep learning model, a spatio-temporal graph convolutional network utilizing gyro-sulcal subdivisions, achieved the highest average classification accuracy in distinguishing resting-state functional magnetic resonance imaging data of methamphetamine use disorders from healthy controls. Within this model, nodes in the lateral orbitofrontal cortex, and the habitual and executive control networks, contributed most significantly to the classification. Additionally, emotional symptom scores were negatively correlated with the sum of negative functional connectivity in the right caudal anterior cingulate sulcus and the functional connectivity between the left putamen and pallidum in methamphetamine use disorders. These findings provide novel insights into the differential functions of gyral and sulcal regions, enhancing our understanding of the neuro-mechanisms underlying methamphetamine use disorders.