A Population-Average, Landmark- and Surface-based (PALS) atlas of human cerebral cortex
Introduction
Atlases play an increasingly important role in providing a spatial and structural framework for visualizing and analyzing many aspects of brain structure, function, and development (Van Essen, 2002, Toga and Thompson, 2002, Mazziotta et al., 2001). In recent decades, numerous human brain atlases have been developed, each with its particular advantages and attractive features but also with various limitations and drawbacks.
For human cerebral cortex, the major challenges in generating a suitable atlas arise from the fact that cortical convolutions are very complex and are highly variable from one individual to the next. To address these challenges, strategic choices must be made on four major fronts: (i) representing cortical morphology and its variability; (ii) visualizing experimental data and underlying cortical structure; (iii) selecting coordinate systems; and (iv) choosing algorithms for registering data from individuals to the atlas. On each of these fronts, a key issue is whether the approach is based on surfaces, volumes, or a combination of both. (i) Representing morphology and individual variability. The core of a brain atlas is a representation of brain morphology derived from one or more individuals, captured by in vivo imaging or postmortem histological methods. Structural magnetic resonance imaging (sMRI) is particularly attractive as a substrate for a human cortical atlas because it can be readily obtained in a population of normal subjects, and it allows for isotropic volume representations that reveal the full complexity of cortical folds. Some human cortical atlases are based on an individual brain or hemisphere, which allows accurate representation of cortical shape in the selected individual but fails to represent variability across the population (Talairach and Tournoux, 1988, Roland et al., 1994, Van Essen and Drury, 1997, Geyer et al., 2001). Other atlases are based on averaging structural MRI volumes across a population of subjects, which capture some shape consistencies across the population, but invariably lead to extensive blurring of cortical convolutions (Evans et al., 1994, Friston et al., 1995, Mazziotta et al., 2001). (ii) Visualization and analysis options. Cortical structure is sufficiently complex that any single visualization mode is inadequate when used in isolation. It is particularly valuable to be able to view atlas data using a combination of volume and surface visualization. (iii) Coordinate systems. Spatial coordinates provide a concise and objective way to represent locations in human cortex. The most basic distinction is between 3D stereotaxic (x,y,z) coordinates and surface-based coordinates (e.g., latitude and longitude on a spherical map). Within each type, numerous specific coordinate systems are in common use, and these differ from one another in important respects. (iv) Registration algorithms. The objective of registering individual brains to an atlas is to compensate for multiple types of variability, including the initial position and orientation of the individual brain, the overall dimensions of each hemisphere, the pattern of cortical convolutions, and the relationship of functional subdivisions (cortical areas) to these convolutions (Toga, 1999). Given the nature of cortical structure and function, it is not possible to remove all variability, i.e., to place all geographic and functional subdivisions into precise correspondence throughout an atlas. Volume-based registration (VBR) algorithms are fundamentally limited because they fail to respect the topology of the cortical sheet and hence can lead to topologically incorrect mappings. Surface-based registration (SBR) avoids this limitation and allows more flexible application of relevant geographic and/or functional constraints, using explicit landmarks (Van Essen et al., 2005) or using metrics that reflect shape characteristics throughout the cortical sheet (Fischl et al., 1999b). However, there are as yet few evaluations of different SBR methods using objective performance measures (Desai et al., 2005).
The present report introduces the Population-Average, Landmark- and Surface-based (PALS) approach and illustrates attractive solutions it offers for each of the aforementioned challenges in generating human cerebral cortical atlases. The specific atlas illustrated here is based on structural MRI volumes from 12 normal subjects and is accordingly identified as the PALS-B12 atlas to distinguish it from other PALS atlases derived from a different set of brains. The PALS-B12 atlas data set includes (i) structural MRI volumes from individual brains and the population average; (ii) cortical segmentations and fiducial surface reconstructions for all contributing hemispheres; (iii) inflated, flattened, and spherical configurations for each hemisphere; (iv) all hemispheres registered to a target atlas surface whose landmarks reflect shape characteristics of the population rather than any individual; (v) individual surfaces resampled to a standard mesh to facilitate simultaneous visualization; (vi) multiple coordinate systems, both stereotaxic and surface-based, for expressing cortical location; (vii) maps of sulcal depth and other shape characteristics for individuals and group averages; (viii) maps of sulcal identity for individuals and the population average; and (ix) many additional types of experimental data mapped from other hemispheres to the PALS-B12 atlas.
Central to the PALS approach is the choice of geographic landmarks used to constrain the registration from individual hemispheres to the atlas. The PALS atlas uses only six well-defined and consistent geographic landmark contours, because other gyral and sulcal features in human cortex are so variable as to be unsuitable (or at least problematic) for use as landmarks. The effectiveness of this approach is demonstrated here using a variety of quantitative analyses related to sulcal shape and sulcal identity. One set of analyses relates to the issue of asymmetries between the left and right hemispheres. Although there is an extensive literature on left–right morphological asymmetries (Toga and Thompson, 2003, Ochiai et al., 2004; see Discussion), much remains unresolved about the nature and magnitude of these differences. The surface-based morphometry (SBM) methods developed in the present study provide valuable additional information regarding left–right asymmetries, especially in the temporal lobe. They are complementary to other SBM analyses (MacDonald et al., 2000, Chung et al., 2003, Salat et al., 2004) that have focused on comparing maps of cortical thickness in different groups rather than the pattern of cortical convolutions.
Surface-based atlases are especially useful for analyzing and comparing complex functional activation patterns obtained in many thousands of fMRI studies involving a wide variety of behavioral tasks and analysis paradigms. Currently, the great majority of fMRI studies obtain statistical power by averaging across subjects using volume-based registration. Mapping such volume-based population-average data onto an individual atlas surface facilitates visualization and comparisons across studies (Van Essen, 2002), but it inevitably introduces significant biases related to the particular convolutions of the target hemisphere. Here, a method for circumventing this bias is introduced, in which fMRI data are mapped separately onto each individual fiducial surface in the PALS-B12 atlas population, then averaged across all maps. This yields an objective representation of complex activation patterns and avoids the bias of a particular individual atlas target.
All of the data sets used in this study are accessible from the SumsDB database (http://sumsdb.wustl.edu:8081/sums). These data can be readily visualized online (using WebCaret) and can be downloaded for offline visualization and analysis (using Caret software). A ‘scene display’ option allows immediate viewing (in either Caret or WebCaret) of the precise screen displays used in generating each of the Caret-derived panels of each figure in this paper.
Section snippets
Methods
Structural MRI volumes were obtained from 24 normal human subjects, all right-handed young adults (ages 18–24; data from Buckner et al., 2004, Head et al., in press). Data from six males and six females were used for the PALS-B12 target atlas. Multiple (three or four) T1-weighted MP-RAGE scans (1.5 T Siemens Vision scanner, 1 × 1 × 1.25 mm voxels) were acquired from each individual. Volumes were registered to the Washington University 711-2C version of stereotaxic atlas space. The 711-2C target
Results
The results are presented in eight sections that address three broad topics. One topic concerns the diversity of cortical shape characteristics; the results can be viewed in a population of individuals (Diversity of sulcal patterns section) or using population-average representations of sulcal depth (Population-based shape representations section) and sulcal identity (Probabilistic surface and volume maps of sulcal identity section). A second topic focuses on methodological considerations,
Discussion
This study was carried out with six major objectives in mind: (i) to introduce the PALS atlas approach as a general way to cope with the complexity of human cortical convolutions and their variability across individuals; (ii) to make the PALS-B12 atlas data set available, including a wide variety of surface and volume data, along with a flexible set of online and offline visualization tools; (iii) to evaluate and compare the quality of surface-based and volume-based registration; (iv) to
Acknowledgments
I thank J. Harwell, D. Hanlon, and J. Dickson for superlative software development; E. Reid for excellent data analysis; S. Danker for manuscript preparation; R. Buckner for MRI data sets; J. Hegde for advice on statistical methods; K. Zilles, K. Amunts, and H. Mohlberg for providing probabilistic architectonic data sets; and A. Snyder, M. Raichle, R. Buckner, and D. Fair for valuable discussions. This project was supported by NIH Research Grant R01-MH-60974, funded by the National Institute of
References (65)
- et al.
Brodmann's areas 17 and 18 brought into stereotaxic space—Where and how variable?
NeuroImage
(2000) - et al.
Analysis of neural mechanisms underlying verbal fluency in cytoarchitectonically defined stereotaxic space—The roles of Brodmann areas 44 and 45
NeuroImage
(2004) - et al.
A unified approach for morphometric and functional data analysis in young, old, and demented adults using automated atlas-based head size normalization: reliability and validation against manual measurement of total intracranial volume
Neuroimage
(2004) - et al.
Deformation-based surface morphometry applied to gray matter deformation
NeuroImage
(2003) - et al.
Volumetric vs. surface-based alignment for localization of auditory cortex activation
Neuroimage
(2005) - et al.
Cortical surface-based analysis: II. inflation, flattening, and a surface-based coordinate system
NeuroImage
(1999) - et al.
Areas 3a, 3b, and 1 of human primary somatosensory cortex
Neuroimage
(1999) - et al.
Areas 3a, 3b, and 1 of human primary somatosensory cortex: Part 2. Spatial normalization to standard anatomical space
Neuroimage
(2000) - et al.
Cerebral asymmetry and the effects of sex and handedness on brain structure: a voxel-based morphometric analysis of 465 normal adult human brains
Neuroimage
(2001) - et al.
Human somatosensory area 2: observer-independent cytoarchitectonic mapping, interindividual variability, and population map
Neuroimage
(2001)
Automated 3-D extraction of inner and outer surfaces of cerebral cortex from MRI
NeuroImage
Human primary auditory cortex: cytoarchitectonic subdivisions and mapping into a spatial reference system
Neuroimage
Sulcal pattern and morphology of the superior temporal sulcus
Neuroimage
Anatomic localization and quantitative analysis of gradient refocused echo-lanaar fMRI suceptibility artifacts
Neuroimage
Probabilistic mapping and volume measurement of human primary auditory cortex
Neuroimage
Automatic recognition of cortical sulci of the human brain using a congregation of neural networks
Med. Image Anal.
New approaches in brain morphometry
Am. J. Geriatr. Psychiatry
Windows on the brain. The emerging role of atlases and databases in neuroscience
Curr. Opin. Neurobiol.
Quantitative analysis of cyto- and receptor architecture of the human brain
Advances in cytoarchitectonic mapping of the human cerebral cortex
Broca's region revisited: cytoarchitecture and intersubject variability
J. Comp. Neurol.
Vergleichende Lokalisationslehre der Grosshirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues
Dissociation of human prefrontal cortical areas across different speech production tasks and gender groups
J. Neurophysiol.
Control of goal-directed and stimulus-driven attention in the brain
Nat. Rev. Neurosci.
Voluntary orienting is dissociated from target detection in human posterior parietal cortex
Nat. Neurosci.
Visual field representations and locations of visual areas V1/2/3 in human visual cortex
J. Vis.
Computerized mappings of the cerebral cortex. A multiresolution flattening method and a surface-based coordinate system
J. Cogn. Neurosci.
Surface-based analyses of the human cerebral cortex
Three-dimensional correlative imaging: applications in human brain mapping
Distributed hierarchical processing in primate cerebral cortex
Cereb. Cortex
High-resolution intersubject averaging and a coordinate system for the cortical surface
Hum. Brain Mapp.
Statistical parametric maps in functional imaging: a general linear approach
Hum. Brain Mapp.
Cited by (982)
SUGAR: Spherical ultrafast graph attention framework for cortical surface registration
2024, Medical Image AnalysisGenetic Influence on Gyral Peaks
2023, NeuroImage