Predicting Cognition from fMRI:A Comparative Study of Graph, Transformer, and Kernel Models Across Task and Rest Conditions

Department of Radiology, Lausanne University Hospital and University of Lausanne (CHUV -UNIL), Lausanne, Switzerland ABSTRACT Predicting cognition from neuroimaging data in healthy individuals offers insights into the neural mechanisms underlying cognitive abilities, with potential applications in precision medicine and early detection of neurological and psychiatric conditions. This study systematically benchmarked classical machine learning (Kernel Ridge Regression) and advanced deep learning models (Graph Neural Networks and Transformer-GNNs) for cognitive prediction using Resting-state, Working Memory, and Language task fMRI data from the Human Connectome Project Y oung Adult (HCP-Y A) dataset. Among the methods compared, a GNN combining structural and functional connectivity consistently achieved the highest performance across all fMRI modalities; however, its advantage over Kernel Ridge Regression using functional connectivity alone was not statistically significant. These findings emphasize the importance of selecting appropriate model architectures and feature representations to fully leverage the spatial and temporal richness of neuroimaging data. This study highlights the potential of multimodal graph-aware deep learning models to combine structural and functional connectivity for cognitive prediction, as well as the promise of Transformer-based approaches for capturing temporal dynamics. By providing a comprehensive comparison of models, this work serves as a guide for advancing brain-behavior modeling using fMRI, structural connectivity and deep learning. INTRODUCTION Understanding and predicting behavior from neuroimaging data in healthy individuals is crucial for advancing our knowledge of the brain's functional architecture and its relationship to behavior. While significant efforts have focused on patients with neurological or psychiatric disorders (Arbabshirani, Plis, Sui, & Calhoun, 2017; Sabuncu, Konukoglu, & Initiative, 2015), the study of healthy participants remains underexplored. Analyzing brain connectivity in healthy individuals can provide valuable insights into the baseline neural mechanisms underlying behavior, offering a foundation for early prognosis of potential neuro or psychiatric conditions (Bassett & Sporns, 2017; Fornito, Zalesky, & Breakspear, 2015; Lui, Zhou, Sweeney, & Gong, 2016; Zhou, Gennatas, Kramer, Miller, & Seeley, 2012). By examining the intricate patterns of functional and structural connectivity, we can identify biomarkers indicative of brain health, which can serve as early indicators of disease susceptibility (M.