FMRI models learn better by organizing tasks and training steps
BrainTaskonomy: Learning How to Pretrain and What to Transfer in fMRI Foundation Models
Computer Vision and Pattern Recognition
Summary
Brain scans using fMRI can be very complex because they come from many different kinds of activities and people. The authors found that treating these different sources and tasks as connected, instead of separate, helps the computer models learn more efficiently. They used a method to figure out which types of brain data to focus on first during training and how to best transfer what is learned to new tasks. This approach improved accuracy in measuring brain activity patterns compared to treating all data equally. Their work shows that understanding the relationships between different brain tasks and training steps helps build better brain models.
fMRIfoundation modelspretrainingtransfer learningcurriculum learningtaskonomybrain imagingmachine learningneuroimagingbudgeted integer programming
Authors
Junfeng Xia, Wenhao Ye, Junxiang Zhang, Jiayu Zuo, Mo Wang, Quanying Liu
Abstract
fMRI foundation models increasingly aggregate heterogeneous data across brain states, cohorts, and acquisition settings, yet pretraining domains are commonly treated as a flat mixture and downstream tasks are adapted independently. We study whether measured learning relations can organize both stages without modifying the backbone. During pretraining, a lightweight Brain-DiT proxy estimates difficulty and directed facilitation across ten fMRI domains, yielding a priority-guided cumulative domain curriculum combined with high-to-low-noise timestep scheduling and joint consolidation. During adaptation, controlled first- and higher-order transfer across fifteen tasks constructs a directed taskonomy, from which budgeted integer programming (BIP) selects directly supervised source tasks and target-specific routes. The joint priority-domain and high-to-low-timestep curriculum reduces v-NMSE, PSD-NMSE, and FC-MSE by 6.5%, 16.3%, and 10.5%, respectively, relative to uniform sampling over both dimensions, and shows strong downstream performance across six in- and out-of-domain tasks. The taskonomy reveals asymmetric, target-dependent transfer, while exploratory sealed-test evaluation shows larger descriptive gains for BIP policies when higher-order route spaces are available than for matched random controls. Together, these findings support organizing fMRI pretraining and adaptation by measured learning relations rather than treating domains and tasks as independent flat sets.