Large Connectome Model: An fMRI Foundation Model of Brain Connectomes Empowered by Brain-Environment Interaction in Multitask Learning Landscape

2025-10-20

AI Summary

Functional neuroimaging suffers from a small-sample bottleneck, severely limiting the clinical generalizability of AI models. To address this, we propose the first foundational model for brain–environment interaction (BEI) based on functional connectomes: it jointly models fMRI data and heterogeneous environmental variables—including behavioral, physiological, and sociodemographic factors—via self-supervised pretraining and multi-task learning, explicitly encoding BEI as tokenized multi-task pretraining objectives. Furthermore, we introduce a pseudo-label–driven semi-supervised fine-tuning strategy to enhance fidelity in modeling disease phenotypes and behavioral outcomes. Evaluated on gender classification, behavioral prediction, and early diagnosis of autism spectrum disorder, Parkinson’s disease, Alzheimer’s disease, and schizophrenia, our model achieves significant gains in cross-task generalization. It establishes a scalable, foundation-model architecture for small-sample neuroimaging analysis.

Technology Category

• Machine Learning: Multimodal Learning
• Humans and AI: Brain-Sensing and Analysis
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

A reliable foundation model of functional neuroimages is critical to promote clinical applications where the performance of current AI models is significantly impeded by a limited sample size. To that end, tremendous efforts have been made to pretraining large models on extensive unlabeled fMRI data using scalable self-supervised learning. Since self-supervision is not necessarily aligned with the brain-to-outcome relationship, most foundation models are suboptimal to the downstream task, such as predicting disease outcomes. By capitalizing on rich environmental variables and demographic data along with an unprecedented amount of functional neuroimages, we form the brain modeling as a multitask learning and present a scalable model architecture for (i) multitask pretraining by tokenizing multiple brain-environment interactions (BEI) and (ii) semi-supervised finetuning by assigning pseudo-labels of pretrained BEI. We have evaluated our foundation model on a variety of applications, including sex prediction, human behavior recognition, and disease early diagnosis of Autism, Parkinson's disease, Alzheimer's disease, and {Schizophrenia}, where promising results indicate the great potential to facilitate current neuroimaging applications in clinical routines.

Problem

• Developing fMRI foundation model using brain-environment interaction
• Improving disease prediction through multitask pretraining approach
• Addressing limited sample size in clinical neuroimaging applications

Innovation

• Multitask pretraining with brain-environment interaction tokens
• Semi-supervised finetuning using pseudo-labels from BEI
• Scalable architecture leveraging demographic and environmental data