Advancing image-based meta-analysis through systematic use of crowdsourced NeuroVault data.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Data Science

Background:

• Image-based meta-analysis (IBMA) synthesizes fMRI study results but faces data accessibility and tool limitations.
• The NeuroVault repository offers a rich source of neuroimaging statistical maps for IBMA.
• Systematic evaluation of NeuroVault's data quality and IBMA methodology is needed.

Purpose of the Study:

• To develop and validate a comprehensive framework for selecting and analyzing neuroimaging data from the NeuroVault repository for IBMA.
• To assess the effectiveness of different image selection strategies (manual vs. heuristic) and meta-analytic estimators.
• To provide an accessible and reproducible methodology for IBMA using publicly available neuroimaging resources.

Main Methods:

• Developed a multi-stage selection framework: preliminary, heuristic, and manual image selection.
• Conducted IBMA for working memory, motor, and emotion processing domains using NeuroVault data.
• Evaluated five meta-analytic estimator methods, focusing on robustness against spurious images.

Main Results:

• IBMA results using manual selection closely matched Human Connectome Project reference maps.
• Heuristic selection methods proved robust, particularly for heterogeneous domains like emotion processing.
• Robust estimators (e.g., median) are crucial for meta-analyses with heterogeneous tasks.

Conclusions:

• The developed framework enables systematic and reproducible IBMA using NeuroVault data.
• Heuristic selection and robust estimators enhance the feasibility and reliability of IBMA, especially with diverse datasets.
• This methodology promotes the use of shared neuroimaging resources, encouraging data sharing and future meta-analyses.