Mapping evolutionary cortical expansion with anatomical MSM

1838 

Abstract Submission 

Burke Rosen¹, Chad Donahue¹, Timothy Coalson¹, John Harwell¹, Renato Besenczi², Emma Robinson², Takuya Hayashi³, David van Essen¹, Matthew Glasser¹

¹Washington University in St. Louis, St. Louis, MO, ²King's College London, London, England, ³RIKEN Center for Biocystems Dynamics Research, Kobe, Hyogo

Burke Rosen  
Washington University in St. Louis
St. Louis, MO

Chad Donahue  
Washington University in St. Louis
St. Louis, MO

Timothy Coalson  
Washington University in St. Louis
St. Louis, MO

John Harwell  
Washington University in St. Louis
St. Louis, MO

Renato Besenczi  
King's College London
London, England

Emma Robinson, Dr  
King's College London
London, England

Takuya Hayashi  
RIKEN Center for Biocystems Dynamics Research
Kobe, Hyogo

David van Essen  
Washington University in St. Louis
St. Louis, MO

Matthew Glasser, Dr.  
Washington University in St. Louis
St. Louis, MO

The surface area of human cortex is ~3-fold greater than that of chimpanzee and >9 times greater than that of the macaque ​​[1]​​. This enlargement is non-uniform across the cortex. For example, primary visual cortex (V1) is only ~2-fold greater in human vs macaque ​​[2]​​. However, previously published maps ​[3–5]​​ of macaque vs human expansion are in poor agreement with one another (Fig. 1A). This may reflect differing registration constraints and algorithms, but one commonality among them is that none have adequately compensated for large areal distortions that arise when projecting anatomical surfaces to the sphere (Fig. 1B). These distortions are particularly deleterious in interspecies registration because the relationship between areal boundaries and cortical folds is very different between ​species ​[1,2]​. To address this issue, we have applied the anatomical Multimodal Surface Matching ​[6]​ (aMSM) algorithm, which was originally designed to enable accurate longitudinal registration of human cortices.

Individuals' cortical surfaces and myelin maps were reconstructed ​[7]​ from high-resolution T1 and T2 weighted MR Images for human (n=1071), chimpanzee (n=19), and macaque (n=32) ​[1,8]​. For each species, a group template surface was created with Connectome Workbench ​[1,7,8]​. Interspecies registration between the template surfaces was performed in two stages. The chimpanzee intermediate used for the first stage served to reduce the necessary areal distortion of each species pair registration. Each species pair registration was performed in both directions and results averaged. To compensate for the loss of surface area introduced by inter-subject averaging of human surface meshes with highly variable folding (Fig. 1C), evolutionary expansion was estimated by resampling each human surface to the macaque mesh and averaging vertex areas across subjects. All registrations were performed using 15 putatively homologous regions (PHRs), or landmarks, as registration features. PHRs were delineated based on previous studies and informed by myelin and other features (Fig. 1D). Parameters were tuned to achieve a balance of feature matching while minimizing areal distortions. Tuning was performed separately for spherical MSM ​[9]​ and aMSM to ensure fair comparison.

We found that interspecies expansion estimated with anatomical MSM resulted in less folding bias than that estimated by spherical MSM (Fig 2A). The application of aMSM results in smaller areal changes in human posterior middle temporal gyrus, an area disproportionately distorted by spherical projection and differences in expansion between the algorithms. The pattern of expansion in chimpanzee vs macaque is quite different from human vs chimpanzee (Fig. 2B). In the former as compared with the latter, expansion is more pronounced in somatomotor areas and more modest in insula. Resampling human individuals then averaging the resulting expansion maps yields the mean evolutionary expansion ratio without a surface averaging loss effect caused by human interindividual folding variability (Fig. 2C). This procedure resulted in smoother expansion maps and approximately 2 times greater absolute expansion when compared to the group average surface used for regularization during aMSM.

Accurate registration of macaque and human cortices is critical for understanding primate brain evolution and for augmenting the clinical and scientific interpretability of animal models. We conclude that using anatomical MSM for interspecies registration and avoiding human surface mesh averaging results in material improvements over prior efforts. Our maps of human vs chimpanzee expansion are consistent with previous findings that a comparatively small somatomotor cortex may be a derived trait in human ​[10]​. Future refinements will incorporate additional features, including myelin maps and fMRI into the registrations.

Image Registration and Computational Anatomy ¹

Cortical Anatomy and Brain Mapping ²

Cortical Cyto- and Myeloarchitecture

Neuroanatomy Other

Cortex

Other - evolution;human;macaque;chimpanzee;MSM;registration;surface;HCP;MRI