Brain simulation augments machine-learning-based classification of dementia.

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Alzheimer's disease (AD) diagnosis relies on complex data.
• Computational brain network modeling offers novel approaches.

Purpose of the Study:

• To enhance Alzheimer's disease (AD) diagnostics using The Virtual Brain (TVB) simulation and machine learning (ML).
• To integrate multi-modal neuroimaging with brain simulation for improved classification accuracy.

Main Methods:

• Utilized TVB for whole-brain simulation, incorporating a cause-and-effect model of amyloid beta (Aβ) and altered excitability.
• Combined Aβ positron emission tomography (PET) and magnetic resonance imaging (MRI) data from 33 ADNI3 participants with simulated local field potentials (LFPs) for ML classification.

Main Results:

• The combined approach significantly improved classification accuracy by approximately 10% (weighted F1-score: 74.28%) compared to empirical data alone (64.34%).
• Identified informative features with high biological plausibility related to AD's spatial distribution.

Conclusions:

• Integrating Aβ-induced hyperexcitation models into ML enhances AD classification.
• Demonstrates TVB's capability in decoding empirical data through connectivity-based brain simulation for AD research.