Predicting the evolution trajectory of population-driven connectional brain templates using recurrent multigraph...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Computational Biology

Background:

• Longitudinal neuroimaging reveals how neurological disorders impact brain structure and function over time.
• Connectional Brain Templates (CBTs) compactly represent population brain multigraphs, integrating multiple connections between brain regions.

Purpose of the Study:

• To predict the temporal evolution of population-level Connectional Brain Templates (CBTs) from baseline data.
• To develop a model for forecasting brain connectivity patterns in both healthy and disordered populations over time.
• To identify potential biomarkers for early diagnosis of neurodegenerative disorders.

Main Methods:

• Introduced the recurrent multigraph integrator network (ReMI-Net), a novel graph neural network architecture.
• Designed ReMI-Net to concurrently learn dependencies at both the multigraph node and time levels.
• Utilized graph convolutional design and normalization layers for accurate template prediction.

Main Results:

• ReMI-Net successfully forecasts well-centered, discriminative, and topologically sound connectional templates over time.
• The model effectively identifies biomarkers differentiating typical and atypical populations.
• ReMI-Net demonstrated superior performance compared to existing benchmarks and state-of-the-art methods in detecting connectivity alterations.

Conclusions:

• ReMI-Net provides a powerful tool for predicting population-level brain connectivity evolution.
• Early identification of neurodegenerative disorder biomarkers is enhanced through this predictive modeling approach.
• The model facilitates new clinical studies by pinpointing critical brain regions and connectivities.