Automating region selection with genetic algorithms for energy landscape analyses of brain dynamics

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Neuroimaging Analysis

Background:

• Understanding brain dynamics is crucial in cognitive neuroscience.
• Energy landscape analysis (ELA) models brain activity but requires manual selection of regions of interest (ROIs).
• Manual ROI selection is subjective and limits objective analysis of brain dynamics.

Purpose of the Study:

• To develop an automated framework, ELA/GAopt, for objective ROI selection in whole-brain atlases.
• To enable data-driven analysis of brain dynamics using genetic algorithms.
• To establish a foundation for biomarker discovery and studying brain state transitions.

Main Methods:

• Developed ELA/GAopt, a framework integrating genetic algorithms with energy landscape analysis.
• Automated the selection of ROIs from whole-brain atlases in a data-driven manner.
• Validated the framework's consistency and reproducibility across three independent datasets.

Main Results:

• ELA/GAopt consistently identified reproducible ROI subsets across datasets.
• The framework detected autism-specific brain dynamics, including global co-activation in sensory-motor and visual networks.
• Replicated findings in multi-site clinical data, demonstrating robustness.

Conclusions:

• ELA/GAopt offers a systematic and objective method for characterizing condition-specific brain dynamics.
• The framework facilitates the exploration of brain state transitions and supports biomarker development.
• This approach enhances the reliability and scalability of neuroimaging analysis for clinical applications.