A new hierarchical Bayesian inference method can detect the community structure of functional brain networks more accurately and reliably than traditional approaches, according to a study published in IEEE Transactions on Medical Imaging by researchers from Monash University and other institutions.
The Research
Led by Lingbin Bian and colleagues, the team developed a multilayer community detection method based on a Bayesian latent block model (LBM). The method estimates the number of communities and their membership in weighted functional networks at both individual and group levels, while preserving the natural variability between subjects. Most existing methods, like modularity maximization, ignore this variability and assume a single community structure for all individuals.
To validate their approach, the researchers first created a synthetic generative model with a known community structure. The Bayesian method recovered the correct community assignments with high consistency. Then they tested it on real working memory task fMRI data from 100 unrelated healthy subjects from the Human Connectome Project (HCP), comparing split-half reproducibility. The new method outperformed commonly used multilayer modularity models in both accuracy and reliability.
Why It Matters
Understanding how brain networks are organized into communities is key to linking brain activity to cognition and behavior. By accounting for individual differences, this method could lead to more personalized insights into cognitive strengths and weaknesses, and potentially help identify biomarkers for neurological conditions. For anyone interested in their own cognitive function, this research underscores that brain network structure is not one-size-fits-all.
What You Can Do
- If you're curious about your own cognitive patterns, consider taking a free, scientifically-validated IQ test to get a baseline.
- Engage in brain training exercises that challenge working memory, as they can help improve network efficiency.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
