Beyond consensus: Embracing heterogeneity in curated neuroimaging meta-analysis
Introduction
Brain imaging experiments are often underpowered (Carp, 2012; Poline et al., 2012; Button et al., 2013). Coordinate-based meta-analysis provides an important framework for analyzing underpowered studies across different experimental conditions and analysis piplines to reveal reliable trends (Wager et al., 2003; Fox et al., 2014; Poldrack and Yarkoni, 2016). Large-scale coordinate-based meta-analyses synthesize thousands of experiments across diverse experimental designs to discover broad and general principles of brain organization and disorder (Laird et al., 2011; Poldrack et al., 2012; Crossley et al., 2014). By contrast, the vast majority of meta-analyses involve smaller number of experiments that are expertly chosen (curated) to generate consensus on specific functional domains (e.g., Binder et al., 2009), brain regions (e.g., Shackman et al., 2011) or disorders (e.g., Cortese et al., 2012).
A popular approach for smaller-scale meta-analyses is activation likelihood estimation or ALE (Turkeltaub et al., 2002; Laird et al., 2005; Eickhoff et al., 2009, 2012; Turkeltaub et al., 2012). ALE identifies brain regions consistently activated across neuroimaging experiments within a functional domain (Costafreda et al., 2008; Spaniol et al., 2009; Beissner et al., 2013) or within a disorder (e.g., Fitzgerald et al., 2008; Minzenberg et al., 2009; Di Martino et al., 2009). Thus, ALE treats heterogeneities across studies as noise. Consequently, ALE analysis might miss out on genuine biological heterogeneity indicative of functional sub-domains or disorder subtypes.
For example, Fig. 1 (middle panel) illustrates activation foci from experiments across multiple tasks associated with a hypothetical functional domain. These foci are generated by two latent sub-domains activating distinct, but overlapping, brain regions. Without prior knowledge of the two sub-domains from theory or previous empirical work, ALE will converge on regions commonly activated across both sub-domains (Fig. 1 left panel). To get around this issue, meta-analytic studies can sub-divide experiments into hypothetical functional sub-domains before applying ALE. For example, a recent meta-analysis divided working memory experiments into verbal versus non-verbal tasks, as well as tasks involving object identity versus object locations (Rottschy et al., 2012). However, manually subdividing experiments requires prior knowledge of the sub-domains and may reinforce biases towards existing concepts. By contrast, in this study, we explored whether a previously published data-driven approach (author-topic model; Yeo et al., 2015) can help uncover heterogeneities1 within ALE-type meta-analyses in a bottom-up, data-driven fashion (Fig. 1 right panel).
A good example in which ALE might miss out on functional sub-domains is the default network and self-generated thought (Smallwood, 2013; Andrews-Hanna et al., 2014). Self-generated thought involves associative and constructive processes that take place within an individual, and depends upon an internal representation to reconstruct or imagine a situation, understand a stimulus, or generate an answer to a question. The term “self-generated thought” serves to contrast with thoughts where the primary referent is based on immediate perceptual input. By virtue of being largely stimulus independent or task unrelated, self-generated thought has been linked with the functions of the default network (Buckner et al., 2008; Andrews-Hanna et al., 2014). Previous ALE meta-analyses have implicated the default network in many tasks involving self-generated thought, including theory of mind, narrative fiction, autobiographical memory and moral cognition (Spreng et al., 2009; Binder et al., 2009; Mar, 2011; Sevinc and Spreng, 2014).
However, many studies have suggested that the default network might be fractionated into sub-systems. For example, Andrews-Hanna and colleagues have proposed a dorsomedial prefrontal subsystem preferentially specialized for social cognition and narrative processing (Andrews-Hanna et al., 2014; Spreng and Andrews-Hanna, 2015) and a medial temporal lobe sub-system preferentially specialized for mnemonic constructive processes (Andrews-Hanna et al., 2014; Christoff et al., 2016). Both sub-systems might spatially overlap or inter-digitate across multiple brain regions (Andrews-Hanna et al., 2014; Braga and Buckner, 2017), which would be challenging to ALE without assuming prior knowledge of the sub-systems (Fig. 1). Furthermore, specific default network fractionation details differed across studies (Laird et al., 2009a; Andrews-Hanna et al., 2010b; Mayer et al., 2010; Humphreys et al., 2015; Kernbach et al., 2018), so application of the author-topic model might potentially clarify sub-systems subserving self-generated thought.
Another common application of ALE is meta-analytic connectivity modeling (MACM), which identifies brain regions that consistently co-activate with a particular seed region (Toro et al., 2008; Koski and Paus, 2000; Robinson et al., 2010; Eickhoff et al., 2010). The assumption is that the seed region exhibits a single co-activation pattern regardless of the actual task activating the seed region (Robinson et al., 2010). However, studies have shown the existence of multiple hub regions in the brain (e.g., dorsal anterior insula, dorsal anterior cingulate cortex) that are activated across many different tasks and might adapt their connectivity pattern depending on task context (Cole et al., 2013; Uddin, 2015; Bertolero et al., 2017). Thus, a seed region might be involved in multiple task-dependent co-activation patterns (McIntosh, 2000).
A good example in which MACM might miss out on multiple co-activation patterns is the left inferior frontal junction (IFJ; Muhle-Karbe et al., 2015). The IFJ has been implicated in many cognitive processes (Brass et al., 2005; Chikazoe et al., 2009a, Chikazoe et al., 2009b; Asplund et al., 2010) and is a key node of the multiple-demand system (Duncan, 2010; Fedorenko et al., 2013). IFJ might also coordinate information among modules by adapting its connectivity patterns across different resting and task states (Cole et al., 2013; Bertolero et al., 2018). Therefore, one might expect the IFJ region to exhibit multiple co-activation patterns that are dependent on task contexts. Since ALE cannot capture heterogeneity across experiments, MACM might be insensitive to such task-dependent co-activation patterns. On the other hand, application of the author-topic model to the IFJ region might yield multiple meaningful co-activation patterns.
In this work, we propose the use of the author-topic model to automatically make sense of heterogeneity within ALE-type meta-analyses. We have previously utilized the author-topic model (Fig. 2; Yeo et al., 2015; Bertolero et al., 2015) to encode the intuitive notion that a behavioral task recruits multiple cognitive components, which are in turn supported by overlapping brain regions (Poldrack, 2006; Leech et al., 2012; Barrett and Satpute, 2013). While our previous work focused on large-scale meta-analysis across many functional domains (Yeo et al., 2015; Bertolero et al., 2015), the current study focuses on heterogeneity within a functional domain (self-generated thought) or co-activation heterogeneity of a seed region (left IFJ). These applications of the author-topic model are made possible by the development of a novel inference algorithm for the author-topic model (Ngo et al., 2016) that is sufficiently robust for smaller-scale meta-analyses.
Our choice of self-generated thought is motivated by previous work suggesting the possibility of fractionating self-generated thought into functional sub-domains (Section 1.1). Similarly, our choice of left IFJ is motivated by previous work suggesting that IFJ might adaptively modify its connectivity patterns across task contexts (Section 1.2). There are of course other functional domains (e.g., executive function) that might be fractionated and other hub regions (e.g., dorsal anterior insula) that might exhibit task-dependent co-activation patterns. Therefore, we have made our code publicly available for researchers to explore the heterogeneity of their preferred functional domain, hub region or mental disorder.
Section snippets
Overview
In Section 2.2, we reviewed the author-topic model and how it could be applied to coordinate-based meta-analysis (Yeo et al., 2015). Section 2.3 discussed simulations and comparisons with spatial independent component analysis. Finally, the model was utilized in two different applications. In the first application (Section 2.4), we applied the author-topic model to discover cognitive components subserving self-generated thought. In the second application (Section 2.5), we estimated the
Overview
In Section 3.2, we show simulation results suggesting that the author-topic model compares favorably with ICA in the goal of discovering latent patterns in coordinate-based meta-analysis. We then explored the cognitive components of self-generated thought (Section 3.3) and the co-activation patterns of the IFJ (Section 3.4). Finally, Section 3.5 discusses a few control analyses.
Simulations
Fig. 5 shows the results of one representative simulation (see Section 2.3 for details). Fig. 5A shows the groundtruth
Discussion
The author-topic model encodes the intuitive notion that behavioral tasks recruited multiple cognitive components, supported by multiple brain regions (Mesulam, 1990; Poldrack, 2006; Barrett and Satpute, 2013). We have previously utilized the author-topic model for large-scale meta-analysis across functional domains (Yeo et al., 2015; Bertolero et al., 2015). By exploiting a recently developed CVB algorithm for the author-topic model (Ngo et al., 2016), we show that the model can also be
Conclusion
Heterogeneities across neuroimaging experiments are often treated as noise in coordinate-based meta-analyses. Here we demonstrate that the author-topic model can be utilized to determine if the heterogeneities can be explained by a small number of latent patterns. In the first application, the author-topic model revealed two overlapping cognitive components subserving self-generated thought. In the second application, the author-topic revealed the participation of the left IFJ in three
Acknowledgements
This work was supported by Singapore MOE Tier 2 (MOE2014-T2-2-016), NUS Strategic Research (DPRT/944/09/14), NUS SOM Aspiration Fund (R185000271720), Singapore NMRC (CBRG/0088/2015), NUS YIA and the Singapore National Research Foundation Fellowship (Class of 2017). Simon Eickhoff is supported by the National Institute of Mental Health (R01-MH074457), the Helmholtz Portfolio Theme “Supercomputing and Modeling for the Human Brain” and the European Union’s Horizon 2020 Research and Innovation
References (160)
- et al.
Machine learning for neuroimaging with scikit-learn
Front. Neuroinf.
(2014) - et al.
Evidence for the default networks role in spontaneous cognition
J. Neurophysiol.
(2010) - et al.
Functional-anatomic fractionation of the brains default network
Neuron
(2010) - et al.
The default network and self-generated thought: component processes, dynamic control, and clinical relevance
Ann. N. Y. Acad. Sci.
(2014) The brain's default network and its adaptive role in internal mentation
Neuroscientist
(2012)- et al.
A central role for the lateral prefrontal cortex in goal-directed and stimulus-driven attention
Nat. Neurosci.
(2010) - et al.
Episodic future thinking
Trends Cognit. Sci.
(2001) - et al.
Back to the future: autobiographical planning and the functionality of mind-wandering
Conscious. Cognit.
(2011) - et al.
The interaction between mood and cognitive function studied with PET
Psychol. Med.
(1997) - et al.
Neural mechanisms of object-based attention
Science
(2014)
Anterior cingulate cortex and response conflict: effects of response modality and processing domain
Cerebr. Cortex
Large-scale brain networks in affective and social neuroscience: towards an integrative functional architecture of the brain
Curr. Opin. Neurobiol.
Probabilistic independent component analysis for functional magnetic resonance imaging
IEEE Trans. Med. Imaging
The autonomic brain: an activation likelihood estimation meta-analysis for central processing of autonomic function
J. Neurosci.
Physiological activation of a cortical network during performance of the Wisconsin Card Sorting Test: a positron emission tomography study
Neuropsychologia
The modular and integrative functional architecture of the human brain
Proc. Natl. Acad. Sci. Unit. States Am.
The diverse club
Nat. Commun.
A mechanistic model of connector hubs, modularity, and cognition
Nature Human Behaviour
Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies
Cerebr. Cortex
The neurobiology of semantic memory
Trends Cognit. Sci.
Parallel interdigitated distributed networks within the individual estimated by intrinsic functional connectivity
Neuron
The role of the inferior frontal junction area in cognitive control
Trends Cognit. Sci.
The role of the frontal cortex in task preparation
Cerebr. Cortex
Anterior cingulate cortex and response conflict: effects of frequency, inhibition and errors
Cerebr. Cortex
Episodic and semantic components of autobiographical memories and imagined future events in post-traumatic stress disorder
Memory
The brains default network
Ann. N. Y. Acad. Sci.
Self-projection and the brain
Trends Cognit. Sci.
The organization of the human cerebellum estimated by intrinsic functional connectivity
J. Neurophysiol.
The human hippocampus and spatial and episodic memory
Neuron
Power failure: why small sample size undermines the reliability of neuroscience
Nat. Rev. Neurosci.
Remembering the past and imagining the future: a neural model of spatial memory and imagery
Psychol. Rev.
Segregation of the human medial prefrontal cortex in social cognition
Front. Hum. Neurosci.
A method for making group inferences from functional MRI data using independent component analysis
Human Brain Mmapping
Neural correlates of the spacing effect in explicit verbal semantic encoding support the deficient-processing theory
Hum. Brain Mapp.
What we talk about when we talk about the default mode network
Front. Hum. Neurosci.
The secret lives of experiments: methods reporting in the fMRI literature
Neuroimage
Functional dissociation in right inferior frontal cortex during performance of go/no-go task
Cerebr. Cortex
Preparation to inhibit a response complements response inhibition during performance of a stop-signal task
J. Neurosci.
Mind-wandering as spontaneous thought: a dynamic framework
Nat. Rev. Neurosci.
Multi-task connectivity reveals flexible hubs for adaptive task control
Nat. Neurosci.
Toward systems neuroscience of ADHD: a meta-analysis of 55 fMRI studies
Am. J. Psychiatry
Predictors of amygdala activation during the processing of emotional stimuli: a meta-analysis of 385 PET and fMRI studies
Brain Res. Rev.
The hubs of the human connectome are generally implicated in the anatomy of brain disorders
Brain
A PET study of cognitive strategies in normal subjects during language tasks: influence of phonetic ambiguity and sequence processing on phoneme monitoring
Brain
Involvement of the inferior frontal junction in cognitive control: meta-analyses of switching and Stroop studies
Hum. Brain Mapp.
Cognitive control in the posterior frontolateral cortex: evidence from common activations in task coordination, interference control, and working memory
Neuroimage
FMRI of past tense processing: the effects of phonological complexity and task difficulty
J. Cogn. Neurosci.
Functional brain correlates of social and nonsocial processes in autism spectrum disorders: an activation likelihood estimation meta-analysis
Biol. Psychiatry
The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour
Trends Cognit. Sci.
The structure of cognition: attentional episodes in mind and brain
Neuron
Cited by (18)
A Functional Magnetic Resonance Imaging Meta-Analysis of Childhood Trauma
2024, Biological Psychiatry: Cognitive Neuroscience and NeuroimagingA transformer-Based neural language model that synthesizes brain activation maps from free-form text queries
2022, Medical Image Analysis