Summary
Healthy aging induces parallel changes in brain functional activity and structural morphology, yet the interplay between these changes remains unclear. Prof. Yuhui Du’s team at the College of Computer and Information Technology, Shanxi University, in collaboration with Prof. Vince D. Calhoun (Georgia State University), analyzed multimodal neuroimaging data from 27,793 healthy subjects (aged 49-76 years) in the UK Biobank. They proposed a unified framework for single-modal and multimodal brain-age prediction and joint functional-structural aging analysis, systematically characterizing diverse synergistic vs. contradictory aging patterns between functional network connectivity (FNC) and gray matter volume (GMV). Importantly, these joint patterns were further linked to specific cognitive decline. The study, titled “Joint aging patterns in brain function and structure revealed using 27,793 samples” was published in Research (2025, 8:0887; DOI: 10.34133/research.0887).

Background
Age-related cognitive decline is closely tied to changes in brain structure and functional interactions. Most prior neuroimaging studies have investigated aging using a single modality, either structural MRI (sMRI) or functional connectivity derived from resting-state fMRI (rs-fMRI). However, brain function and structure are mutually coupled and co-develop during aging. Examining only one modality makes it difficult to disentangle the true mechanisms of cognitive aging. Moreover, many multimodal brain-age studies simply concatenate functional and structural features, which often allows the stronger structural features to overshadow more subtle functional contributions. As a result, genuine joint aging patterns may be missed.

Research Progress
To address these challenges, the authors developed a unified multimodal brain-age prediction and joint aging analysis framework (Fig. 1). Specifically, age was predicted separately from FNC and from whole-brain GMV using nested double-layer 10-fold cross-validated Lasso regression, yielding stable age-related functional and structural features. Important FNC and GMV features identified from single modalities were fused and evaluated under the same nested cross-validation scheme, ensuring a fair multimodal comparison and preventing structural dominance. Finally, each reliable FNC was paired with the GMVs of its two connected regions to form joint aging changes, enabling systematic characterization of synergistic (or contradictory) functional-structural changes and their cognitive relevance.

The study confirmed that GMV-based models outperformed FNC-based models in age prediction, indicating stronger structural sensitivity to aging. Critically, the multimodal model combining FNC and GMV features achieved the highest predictive accuracy (Fig. 2), underscoring the necessity of integrated analysis for a comprehensive understanding of brain aging.

Further analysis of joint FNC-GMV changes revealed two primary aging patterns (Fig. 3):
Synergistic changes: Concurrent decline in FNC strength and GMV, predominantly observed in the cerebellum, frontal pole, paracingulate gyrus, and precuneus cortex. This pattern indicates coordinated functional and structural degeneration in regions governing motor control and higher-order cognition.

Contradictory changes: Increased FNC coupled with GMV reduction, primarily occurring in visual areas like the occipital pole and lateral occipital cortex. This suggests adaptive functional enhancement to counteract structural decline.

Notably, specific joint changes were strongly associated with declines in distinct cognitive domains (Fig. 4). Contradictory changes in visual areas correlated most strongly with fluid intelligence and numeric memory, reflecting adaptive maintenance of visual-information processing. In contrast, synergistic decline between the cerebellum Crus I and paracingulate gyrus was linked to slower reaction time, indicating direct impacts of sensorimotor and attention circuit degradation.

Significance and Future Prospects
This large-scale study provides direct evidence of joint functional-structural changes in healthy brain aging, revealing a complex dynamic process involving both widespread synergistic degeneration and localized compensatory adaptation. The findings not only advance our understanding of the neurobiological mechanisms underlying differential cognitive decline but also lay a foundation for developing multimodal neuroimaging biomarkers and targeted early-intervention strategies.

The complete study is accessible via DOI:10.34133/research.0887