Cortical Folding Shows Fingerprinting Ability in Early Developing Rhesus Macaques

2123 

Abstract Submission 

Yilan Yin¹, Zhengwang Wu¹, Weili Lin¹, Li Wang¹, Gang Li¹

¹University of North Carolina at Chapel Hill, Chapel Hill, NC

Yilan Yin  
University of North Carolina at Chapel Hill
Chapel Hill, NC

Zhengwang Wu  
University of North Carolina at Chapel Hill
Chapel Hill, NC

Weili Lin  
University of North Carolina at Chapel Hill
Chapel Hill, NC

Li Wang  
University of North Carolina at Chapel Hill
Chapel Hill, NC

Gang Li  
University of North Carolina at Chapel Hill
Chapel Hill, NC

Previous human studies have revealed that complex cortical folding patterns have the fingerprinting ability for individual identification as early as 30 postmenstrual weeks and remain a stable individual identifier across ages (Duan et al., 2020). However, it is unknown if this fingerprinting ability can extend to nonhuman primates with simpler cortical folds, especially during early brain development marked by dramatic cortical development. For the first time, we perform individual identification tasks based on a longitudinal dataset including 156 rhesus macaque scans during early postnatal stages. Our results have shown 100% identification accuracy, revealing the fingerprinting ability of cortical folds in macaques from birth.

In total, 156 longitudinal MRI scans of 32 rhesus macaques (18 males) from the UNC-Wisconsin Rhesus Macaque Neurodevelopment Database are used in this study (Young et al., 2017). Each macaque subject has 4 to 5 scans ranging from 0 to 3 years in age as illustrated in Fig. 1. (a). To characterize the cortical folding, we first segment the macaque brain MRI into gray matter, white matter, and cerebrospinal fluid; reconstruct cortical surfaces (Li et al., 2012); and compute the mean curvature, average convexity, and sulcal depth (Li et al., 2014) to vertex-wisely chart the cortical folding pattern. Then, we align all reconstructed cortical surfaces into a common space using an unbiased longitudinal group-wise registration strategy. (Li et al., 2015). The cortical lobar parcellation (10 regions in each hemisphere) is propagated from an atlas (Styner et al., 2007) to each individual scan. Finally, we resample all aligned cortical surfaces with 40,962 vertices to establish vertex-wise correspondence across scans and subjects.

The identification task uses global-based, ROI-based, and vertex-wise frameworks in two directions: identifying a subject's later scan using an earlier one, and vice versa. A dynamic identification pool is formed by combining a subject's two scans at different times with the scans of all other subjects. The match is chosen by ranking pair-wise Pearson correlations using a specific feature combination. In the global-based framework, each scan is represented by concatenated features from both hemispheres, followed by subject-wise Pearson correlation ranking (Duan et al., 2020). In the ROI and vertex-wise frameworks, the selection process is performed at each ROI or within a 14-ring neighborhood from uniformly distributed vertices on both hemispheres (excluding noncortical areas), resulting in 1 candidate at each region or vertex. Then, a voting strategy is employed to count the votes for the candidate selected at each region or vertex, leading to the final decision of the match (Duan et al., 2020). The identification accuracy at each cortical region or vertex is calculated as the probability of correctly identifying the match. The vertex-wise accuracy map is averaged for each vertex, which is used repeatedly in defining neighborhood regions.

Fig. 1 shows the results for different cortical feature combinations: (a) scan age distribution for each subject; (b) sulcal depth maps of human infants at 41 postmenstrual weeks and macaques at 20 months; and (c) the identification accuracies for both forward (early to late) and backward (late to early) tasks. Fig. 2 displays the maps of identification accuracy based on (a) 14-ring neighborhoods of around 2,300 uniformly distributed vertices and (b) lobar ROI for mean curvature, average convexity, and sulcal depth in the forward task. Results indicate high identification accuracy in the superior and middle temporal gyri for all cortical features.

This study unprecedently reveals that cortical folding patterns are reliable individual markers of rhesus macaques with high fingerprinting ability during dynamic early postnatal brain development, despite that macaques have simpler and less individualized cortical shapes than humans.

Cortical Anatomy and Brain Mapping ¹

Neuroanatomy Other

Brain Atlases

Informatics Other ²

Anatomical MRI

Cortex

Development

Informatics

Morphometrics

MRI