Enhancing task fMRI individual difference research with neural signatures

Task-based functional magnetic resonance imaging (tb-fMRI) has advanced our understanding of brain-behavior relationships. Standard tb-fMRI analyses suffer from limited reliability and low effect sizes, and machine learning (ML) approaches often require thousands of subjects, restricting their ability to inform how brain function may arise from and contribute to individual differences. Using data from 9,024 early adolescents, we derived a classifier ('neural signature') distinguishing between high and low working memory loads in an emotional n-back fMRI task, which captures individual differences in the separability of activation to the two task conditions. Signature predictions were more reliable and had stronger associations with task performance, cognition, and psychopathology than standard estimates of regional brain activation. Further, the signature was more sensitive to psychopathology associations and required a smaller training sample (N=320) than standard ML approaches. Neural signatures hold tremendous promise for enhancing the informativeness of tb-fMRI individual differences research and revitalizing its use.