Neuroimaging meta-analysis for consensus and discovery
Project status: Complete
Data-driven meta-analytic clustering
Abstract
Although most often used to assess agreement or convergence across published studies, neuroimaging meta-analysis is a powerful tool for uncovering disparate, recurrent patterns of brain activation from the literature. By applying unsupervised clustering approaches to a dataset of 3D stereotaxic brain activation coordinates, we uncovered multiple activation patterns underlying two very different concepts in the neuroimaging literature: naturalistic fMRI paradigms and reward processing studies. Naturalistic paradigms employ dynamic stimuli that demanded continuous, real-time integration of dynamic streams of information and have been used to study a wide range of behaviors and cognitive concepts. Across 110 neuroimaging papers, we uncovered 6 recurrent patterns of brain activation, related to sensory input, attentional control, and domain-specific processing (i.e., language, affect, and spatial processing). On the other hand, reward processing is a more cohesive domain, with various external influences and internal facets. Here, meta-analytic clustering identified seven recurrent patterns of brain activation from 176 published neuroimaging studies. These patterns reflect specialized processing involved in predicting value and processing emotional, external, and internal influences, concurrent with accepted reward learning theories.
Manuscripts
Presentations
Theory-driven meta-analysis for domain-specific processing
Neuroimaging meta-analysis is often applied in a more theory-driven manner to assess nuances in brain activation related to different facets of specific cognitive domains. Across a domain (e.g., social processing, problem solving), differences in experimental paradigm or statistical contrasts between studies can uncover different aspects of neural processing that would otherwise be lumped together in the literature or seemingly unrelated to other papers due to authors' phrasing. However, manual curation of meta-analytic corpora can be applied to apply domain-specific knowlege to group like papers according to underlying theory (e.g., the NIH Research Domain Criteria, RDoC, for social processing).
Manuscripts
Functional neuroanatomy via meta-analysis
By selecting published papers for a meta-analysis not by the paper's topic or methods, but by the brain activation coordinates reported in the paper, meta-analysis can be used to assess functional neuroanatomy. For example, selecting papers that report coordinates within a region (e.g., hippocampus, hypothalamus) and meta-analyzing their results can provide insight into that region's functional connections throughout the brain (i.e., meta-analytic coactivation mapping, MACM) and its intrinsic functional organization (i.e., meta-analytic coactivation-based parcellation, CBP).
My undergraduate research mentor, Jennifer Robinson, and used meta-analytic coactivation-based parcellation, in addition to resting-state functional and diffusion-weighted magnetic resonance imaging at 7T to delineate the functional topography of the human hippocampus. In my undergraduate work, I applied meta-analytic coactivation mapping to assess the functional connectivity of the human hypothalamus.