MULAN: Evaluation and ensemble statistical inference for functional connectivity.

Area of Science:

• Neuroscience
• Computational Biology
• Data Science

Background:

• Extracting functional connectivity graphs from neuronal networks is crucial for understanding brain function.
• Current analysis methods lack confidence due to variability in results, parameter sensitivity, and insufficient systematic evaluation.

Purpose of the Study:

• To introduce MULAN (MULtiple method ANalysis), an ensemble-based approach to enhance the reliability of functional connectivity graph extraction.
• To reduce dependency on parameter settings and provide a direct estimate of the most likely connectivity graph with confidence levels.

Main Methods:

• MULAN combines multiple analysis methods with fuzzy logic for robust graph extraction.
• A genetic algorithm optimizes parameter settings using simulated datasets mirroring experimental temporal structures.
• The approach includes validation steps and cross-validation of experimental data subsets.

Main Results:

• MULAN successfully extracts functional connectivity graphs with improved confidence.
• Parameter optimization via genetic algorithms reduces sensitivity to individual settings.
• Systematic evaluation validates the strategy against stereotactic electroencephalography (SEEG) and functional magnetic resonance imaging (fMRI) data.

Conclusions:

• MULAN offers a more reliable method for functional connectivity analysis in neuronal networks.
• The ensemble approach and optimized parameter selection enhance the accuracy and trustworthiness of extracted connectivity graphs.
• This strategy provides a robust framework for analyzing complex neuroimaging data.