Individual parcellations defined from connectivity improve heritability of cortical thickness
Poster No:
2001
Submission Type:
Abstract Submission
Authors:
Clément Langlet1, Denis Rivière1, Antoine Grigis1, Vinecnt Frouin1, Jean-François Mangin1
Institutions:
1Université Paris-Saclay, CEA, CNRS, Neurospin, Gif-sur-Yvette, France
First Author:
Co-Author(s):
Introduction:
Current studies often consider surface phenotypes defined from regions stemming from an atlas projected on individuals via spatial normalization driven by the curvature of the largest cortical folds. This strategy is suboptimal relative to interindividual variations of the spatial organization of the architectural entities making up the cortical surface (particular region topography (Glasser 2016) or rare folding architecture (Mangin 2019)). To propose an alternative, we used a structural-connectivity-based subdivision (Lefranc 2016) - Constellation atlas - of the Desikan atlas (Desikan 2006) further projected on 1004 individuals (Langlet 2023) of the Human Connectome Project dataset (Van Essen 2013) according to their structural connectivity fingerprint. Here, we assessed the relevance of these individual parcellations by comparing them with 1000 randomly generated subdivisions of the Desikan atlas on the regional cortical thickness phenotype. We then performed a heritability study on the regional cortical thickness phenotype defined from the individual parcellations.
Methods:
The 1000 randomly generated parcellations were obtained using Voronoi diagrams with the same granularity as the Constellation group atlas and act as group atlases.
Firstly, we separately aggregated the thickness standard deviations of Constellation-based parcels and Voronois-based parcels and performed a Student T-test under the null hypothesis : {Given a Desikan region, Constellation subdivisions have a lesser thickness standard deviation than the ones defined from Voronoi parcellations}. In significant regions, the Constellation subdivisions provide better estimation of the regional thickness phenotype than random splits with similar sizes.
Secondly, in these significant regions, we studied how the regional thickness of Constellation parcels relates to the ones yielded by the Voronoi parcellations. As regions cannot be matched across template parcellations, we paired their regional thicknesses on a vertex basis. We then averaged each regional thickness across the 1004 subjects and computed the Z-scores of the regional thickness stemming from Constellation subdivisions in the distribution yielded by the 1000 Voronoi subdivisions.
Finally, we performed a twin-based heritability study of the regional thickness phenotype defined from Constellation individual parcellations, using the SOLAR algorithm (Almasy 1998). As covariates, we took into account the effect of age, age², sex and age × sex. We thus obtained heritability scores for the Constellation subdivisions and Desikan base regions. We kept scores that remained significant after Bonferroni correction for the number of regions of each parcellation scheme.
Firstly, we separately aggregated the thickness standard deviations of Constellation-based parcels and Voronois-based parcels and performed a Student T-test under the null hypothesis : {Given a Desikan region, Constellation subdivisions have a lesser thickness standard deviation than the ones defined from Voronoi parcellations}. In significant regions, the Constellation subdivisions provide better estimation of the regional thickness phenotype than random splits with similar sizes.
Secondly, in these significant regions, we studied how the regional thickness of Constellation parcels relates to the ones yielded by the Voronoi parcellations. As regions cannot be matched across template parcellations, we paired their regional thicknesses on a vertex basis. We then averaged each regional thickness across the 1004 subjects and computed the Z-scores of the regional thickness stemming from Constellation subdivisions in the distribution yielded by the 1000 Voronoi subdivisions.
Finally, we performed a twin-based heritability study of the regional thickness phenotype defined from Constellation individual parcellations, using the SOLAR algorithm (Almasy 1998). As covariates, we took into account the effect of age, age², sex and age × sex. We thus obtained heritability scores for the Constellation subdivisions and Desikan base regions. We kept scores that remained significant after Bonferroni correction for the number of regions of each parcellation scheme.
Results:
The selection process using thickness standard deviation yielded a total of 23 significant regions across the two hemispheres. In these 23 regions, Figure 1 shows how the regional thickness differs for the Constellation method compared with random subdivisions. We observe that the individual parcellations better segregate areas with extreme cortical thicknesses thus possibly separating different structural entities. We show results of the heritability study in Figure 2 and observe that, by using individual parcellations, we tend to focus the heritability score of the Desikan regions on one subdivision. This even performed better in some regions such as the left paracentral: the heritability score of the Desikan region is 0.60 whereas one subdivision has an heritability score of 0.65.
Conclusions:
To conclude, we argue that the use of individual parcellations better captured structural entities than would random parcellations. The heritability study confirmed this result especially in associative areas, as heritability scores were focused on one sub-region. In a recent study, highly connected regions were found more heritable (Arnatkeviciute 2021) - in line with the results proposed here - hence these individual parcellations could be of use for further genetic studies.
Genetics:
Genetic Modeling and Analysis Methods 2
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Diffusion MRI Modeling and Analysis
Segmentation and Parcellation 1
Neuroinformatics and Data Sharing:
Brain Atlases
Keywords:
Atlasing
Cortex
MRI
Tractography
White Matter
Other - parcellations; intersubject variability;heritability;cortical thickness;statistical analysis
1|2Indicates the priority used for review
Provide references using author date format
Arnatkeviciute, A. (2021), 'Genetic influences on hub connectivity of the human connectome', Nature Communications, 12.
Desikan, R. S. (2006), 'An automated labeling system for subdividing the human cerebral cortex on mri scans into gyral based regions of interest', NeuroImage, 31.
Glasser, M. F. (2016), 'A multi-modal parcellation of human cerebral cortex', Nature, 536(7615):171–178.
Langlet, C. (2023), 'Nested parcellations connectome delivered for one large dataset using constellation algorithm (v1.2)', Ebrains.
Lefranc, S. and Roca, P. (2016), 'Groupwise connectivity-based parcellation of the whole human cortical surface using watershed-driven dimension reduction', Medical Image Analysis, 30.
Mangin, J.-F. (2019), '“Plis de passage” Deserve a role in models of the cortical folding process', Brain Topography, 32.
Van Essen, D. C. (2013), 'The WU-Minn Human Connectome Project: An overview', NeuroImage, 80:62–79.