Cerebellar meta-analytic maps: describing bias to construct domain-specific functional topographies

1347 

Abstract Submission 

Neville Magielse^1,2,3, Amin Saberi¹, Aikaterina Manoli², Simon Eickhoff¹, Sofie Valk²

¹INM-7, Research Center Jülich, Jülich, Germany, ²Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ³Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf, Düsseldorf, Germany

Neville Magielse  
INM-7, Research Center Jülich|Max Planck Institute for Human Cognitive and Brain Sciences|Institute of Systems Neuroscience, Heinrich Heine University Düsseldorf
Jülich, Germany|Leipzig, Germany|Düsseldorf, Germany

Amin Saberi  
INM-7, Research Center Jülich
Jülich, Germany

Aikaterina Manoli  
Max Planck Institute for Human Cognitive and Brain Sciences
Leipzig, Germany

Simon Eickhoff  
INM-7, Research Center Jülich
Jülich, Germany

Sofie Valk  
Max Planck Institute for Human Cognitive and Brain Sciences
Leipzig, Germany

The cerebellum, apart from its canonical motor functions, plays important roles in most human behaviors, including language, theory of mind, working memory, and executive functioning (King et al. 2019). Cerebellar activations associated with these diverse behavioral domains have been reported in the literature (Kruithof, Klaus, and Schutter 2023; Van Overwalle, D'aes, and Mariën 2015; Stoodley 2012; Stoodley and Schmahmann 2009). However, a full cerebellar functional topography based on the accumulation of neuroimaging studies is still missing. Here, we aimed to construct such meta-analytic maps of cerebellar activity in a cross-domain manner, leveraging the BrainMap database. Together, we show both promise and challenge in consolidating cerebellar functional imaging data into cross-domain meta-analytic maps.

We used the BrainMap annotated database of the neuroimaging literature to identify task-based functional imaging experiments on healthy subjects that reported cerebellar activations (Laird et al. 2009; Fox et al. 2005). We categorized the experiments based on the five main behavioral domains (BDs) in the BrainMap including action, cognition, emotion, interoception, and perception. In each domain we performed activation likelihood estimation (ALE) to investigate convergence of findings within the cerebellar mask (Eickhoff et al. 2012). Additionally, we performed a domain-general ALE to assess whether there is spatial bias in the reported findings within the cerebellum. In this analysis we included 100 randomly selected experiments from each BD to ensure even participation of the domains. Subsequently, we reported the ALE maps and evaluated their spatial cross-correlation. Last, we studied the co-alignment of the ALE maps with an openly available atlas of cerebellar activations associated with the multi-domain task battery (King et al. 2019).

We found a total of 1679 experiments reporting cerebellar task-associated activations. The ALE maps of the cognition, emotion and perception BDs were highly similar, while the ALE map of action was most similar to perception, and the ALE map of interoception showed the least similarity to the other domains (Fig. 1A, B). There was a domain-general convergence of cerebellar activations in primarily the antero-superior cerebellum, potentially indicating a spatial bias in the reported findings. Accordingly, we found that the distribution of findings across axial slices was unequal across the cerebellar volume (Fig. 1C). Zooming in, we observed positive correlations between the maps of 2-back and math tasks with the cognitive BD ALE, and finger sequence task with the action BD ALE map, while the correlation of resting-state and movie watching tasks with the BD ALE maps were generally negative (Fig. 2).

Together, our study shows a meta-analytic mapping of task activations in the cerebellum which extends previous studies done at the level of individuals (King et al. 2019). We observed a bias of reported findings in the anterior-superior cerebellum. Common challenges of cerebellar neuroimaging, including BOLD-bleeding of visual cortical signal, heterogeneous signal-to-noise, and primarily a severe sampling bias, should be addressed in future neuroimaging and meta-analytic studies.

Higher Cognitive Functions Other

Activation (eg. BOLD task-fMRI) ¹

Brain Atlases ²

Workflows

Cerebellum

Meta- Analysis