Individual Differential Functional Brain Networks Based on Sparsity and Dissimilarity

Individuals exhibit substantial variability in behavior and phenotypic traits, presenting a key challenge in neuroscience when attempting to link these inter-individual differences to the functional properties of specific brain regions. However, previous functional connectivity (FC)-based prediction studies have primarily focused on building predictive models using individual FCs without considering the individual differences between subjects. Here, we propose a sparsity and dissimilarity-based individual differential functional brain network (SDIDFBN). This approach captures individual differential FCs by sparsifying the subject-level differences while incorporating global-level dissimilarities through cosine distance. Results show that SDIDFBN consistently achieve the highest accuracy in predicting symptoms and cognition for schizophrenia, compared with 4 brain network construction methods. Moreover, SDIDFBN successfully distinguishes schizophrenia patients from healthy controls across four classifiers in two independent datasets. Overall, SDIDFBN effectively and reliably captures individual differential functional brain network features for both prediction and classification tasks, demonstrating its potential in identifying individualized brain networks in personalized prediction and diagnosis of brain disorders.Clinical Relevance- SDIDFBN effectively extract individual differences in functional brain network features, which provides potential clinical applications for individualized diagnosis of brain disorders.