桥接(联网)
计算机科学
连接体
人工智能
神经科学
神经影像学
神经生理学
医学影像学
大脑定位
模式识别(心理学)
计算机视觉
人类连接体项目
可视化
作者
Jing Zhang,Li Su,Rong Zhang,Tianming Liu,Dajiang Zhu
标识
DOI:10.1109/tmi.2026.3755232
摘要
Integrating brain imaging data with clinical reports enables the effective use of complementary multimodal information, supporting more accurate and timely diagnosis in real-world clinical settings. This approach has gained growing attention in brain disorder research, yet significant limitations remain: how to encode brain structure and function to reflect meaningful neurological patterns, and how to effectively link objective imaging data with subjective text-based reports while establishing interpretable associations. In this work, we propose a novel multimodal framework that aligns brain connectomes with clinical reports in a shared cross-modal latent space at both subject and connectome levels. Our key innovation lies in treating brain subnetworks as tokens rather than raw image patches, enabling more efficient identification of system-level associations between neuroimaging findings and clinical observations. These cross-modal associations are further fused to enhance mild cognitive impairment (MCI) prediction and provide interpretable diagnostic insights. To our knowledge, this is the first attempt to integrate both structural and functional brain connectivity with non-imaging clinical information. Experimental validation on the ADNI dataset demonstrates performance competitive with recently reported methods. Notably, despite no explicit connectome-token-level supervision, the learned connectome-text associations recover multiple relationships independently reported in the neuroimaging and clinical literature, suggesting that the framework captures clinically coherent cross-modal associations. This pioneering work advances multimodal artificial intelligence (AI) for neurodegenerative disease diagnosis by bridging the gap between neuroimaging analysis and clinical text interpretation through innovative connectome tokenization and cross-modal alignment strategies.
科研通智能强力驱动
Strongly Powered by AbleSci AI