Functional fingerprinting for the developing brain using deep metric learning: an ABCD study

ABSTRACT Brain fingerprinting is a promising approach for characterizing the uniqueness of individual brain functioning using functional magnetic resonance imaging (fMRI) data. Here, we propose a novel deep learning framework, the metric-BoIT, for brain fingerprinting and demonstrate its effectiveness in capturing individual variability among early adolescents undergoing dramatic brain changes. Utilizing resting-state fMRI data from the Adolescent Brain Cognitive Development (ABCD) dataset, we identified brain functional fingerprints that achieved remarkable individual identification accuracy, reaching 97.6% within a single session and maintaining 86.6% accuracy over a four-year developmental period. Annotation analysis revealed that higher-order association regions, particularly those within the default-mode network, contributed most significantly to these distinctive brain fingerprints ( t = 5.618, p < 0.001). Moreover, these brain fingerprints were relevant to cognitive functions, as evidenced by significant correlations with fluid intelligence ( F = 1.282, p = 0.027) and crystallized intelligence ( F = 1.405, p < 0.001). The extracted brain fingerprints were additionally associated with genetics, showing that individuals with strong genomic relationships exhibited more similar brain fingerprint patterns ( t = −12.330, p < 0.001). Together, our study not only presents an innovative approach to brain functional fingerprinting but also provides valuable insights into the individual variability underlying adolescent neurodevelopment.