Revealing Fine-grained Genetically Informed Parcellation Maps of Neonatal Cerebral Cortex

2304 

Abstract Submission 

Ying Huang¹, Zhengwang Wu², Tengfei Li², Ya Wang², Xifeng Wang², Hongtu Zhu², Weili Lin², Li Wang², Jun Feng¹, John Gilmore³, Gang Li²

¹Northwest University, Xi'an, ShaanXi, ²University of North Carolina at Chapel Hill, Chapel Hill, NC, ³University of North Carolina Chapel Hill, Chapel Hill, NC

Ying Huang  
Northwest University
Xi'an, ShaanXi

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

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

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

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

Hongtu Zhu  
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

Jun Feng  
Northwest University
Xi'an, ShaanXi

John Gilmore  
University of North Carolina Chapel Hill
Chapel Hill, NC

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

Genetic factors have been proven to be one of the major determinants in shaping the neonatal cerebral cortex (Huang et al., 2023; Jha et al., 2018). Prior research has demonstrated distinct genetic influences on the spatial patterns of cortical properties, like cortical thickness (CT) and surface area (SA) in neonates, leading to their unique genetically informed parcellation maps (Huang et al., 2023). However, these parcellation maps were derived with coarse scales and based on single cortical properties, making them unable to comprehensively characterize the fine-grained genetically regulated patterns of the neonatal cerebral cortex. To fill this knowledge gap, by combining genetic correlations from multiple cortical properties, i.e., CT and SA, we aimed to reveal a joint, fine-grained, genetically informed parcellation map of the neonatal cerebral cortex through a multi-view spectral clustering approach (Kumar et al., 2011).

T1-weighted and T2-weighted structural MR images from 202 same-sex twin neonates were adopted in this study (Jha et al., 2018). Cortical surfaces were reconstructed, aligned, and resampled using an infant-dedicated computational pipeline, iBEAT V2.0 (Li et al., 2014; Li et al., 2015; Li et al., 2019; Wang et al., 2018). After computing CT and SA, the genetic correlations of CT (SA) between any two cortical vertices were calculated using the standard bivariate ACE twin model (Neale and Cardon, 2013). Then, with the obtained genetic correlation matrices of CT and SA, we performed a multi-view spectral clustering method (Kumar et al., 2011) to group all cortical vertices into k distinct regions, with each region consisting of a group of vertices that are strongly genetically correlated. Finally, to assess the clustering performance and choose the appropriate region number, three criteria were employed, namely silhouette coefficients (Rousseeuw, 1987), adjusted Rand index (ARI) (Hubert and Arabie, 1985), and normalized mutual information (NMI) (Strehl and Ghosh, 2002). Higher values of these criteria generally indicate better clustering performance. Of note, the ground truth utilized in computing both ARI and NMI was obtained based on the intersection of prior genetic parcellation maps of neonatal CT and SA (Huang et al., 2023).

Fig. 1 illustrates the evaluation results derived from the three criteria. As can be seen, the local maxima for all three criteria coincide at region numbers 35, 67, and 86. Considering the spatial distribution and hemispheric symmetry of parcellation maps, we chose 86 as the appropriate region num.
Furthermore, to enhance the anatomical comparability of corresponding regions in both hemispheres, bilaterally symmetric regions were manually combined and assigned shared IDs. After the aforementioned processes, we obtained the final fine-grained, genetically informed parcellation maps of the neonatal cerebral cortex, with each hemisphere consisting of 36 regions. As depicted in Fig. 2, the parcellation maps exhibit bilaterally symmetric patterns and correspond well to structurally or functionally meaningful areas. Besides, in accordance with the numbers labeled on figure, the approximate names of these regions are shown below the parcellation maps.

Leveraging neuroimages of neonatal twins, for the first time, we revealed the fine-grained, genetically informed parcellation maps of the neonatal cerebral cortex based on multiple cortical properties, i.e., CT and SA. The discovered parcellation maps comprehensively reflect genetically regulated detailed patterns of the neonatal cerebral cortex and are structurally and functionally meaningful.

Genetic Association Studies ²

Early life, Adolescence, Aging

Cortical Anatomy and Brain Mapping

Anatomical MRI ¹

Cortex

Development

STRUCTURAL MRI