Meta-analysis of the functional neuroimaging literature with probabilistic logic programming

Area of Science:

• Cognitive Neuroscience
• Artificial Intelligence
• Neuroimaging

Background:

• Synthesizing evidence from functional neuroimaging studies is crucial for reliable brain-behavior association inference.
• Current meta-analysis tools are limited in the complexity of neuroscience concepts and questions they can address.

Approach:

• We introduce NeuroLang, a domain-specific language utilizing probabilistic first-order logic programming for neuroimaging meta-analysis.
• NeuroLang integrates formalisms from artificial intelligence and knowledge representation to enhance hypothesis expression and testing.
• This approach effectively models the inherent uncertainty in neuroimaging data.

Key Points:

• NeuroLang significantly expands the scope of questions addressable in neuroimaging meta-analyses.
• The language facilitates the investigation of complex structure-function associations.
• Demonstrated use cases include inferring roles of canonical brain networks and analyzing the frontoparietal control network's organization.

Conclusions:

• NeuroLang significantly expands the expressivity and capability of neuroimaging meta-analysis.
• The language effectively models data uncertainty and addresses a wider range of scientific questions.
• NeuroLang facilitates deeper insights into brain-behavior associations and network organization.