• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Neuroimage. Author manuscript; available in PMC Oct 1, 2008.
Published in final edited form as:
PMCID: PMC2227945
NIHMSID: NIHMS31196

What is Where and Why it is Important

The paper by Devlin and Poldrack entitled ‘In praise of Tedious Anatomy’ deals with an increasingly important topic. Namely, how do we effectively communicate, with appropriate accuracy and precision, the ‘where’ of our measurements? They make cogent arguments for the use of anatomical maps to reference the location of functionally active regions of brain as measured with fMRI. We agree with this premise and would like to amplify further (with some additional suggestions) the rationale for equating functional measures with observations about structure. But before we do that, it might be useful to first generalize this issue, as one that pertains to 1) comparing data collected across multiple subjects, modalities, experiments and laboratories and 2) integrating this data as a comprehensive whole.

Specifying the site of activation, in anatomic terms, is intended (in part) to provide the ability to compare and contrast different studies. The notion of an atlas is, among other things, to help with this localization. Unfortunately many anatomic terms are sometimes interpreted variably and the boundaries of anatomic regions are sometimes in dispute. Given the mostly descriptive history of anatomy, its application, in a traditional sense, to modern volumetric MRI surveys of the brain sometimes produces more problems than it solves. Atlases such as the (Talairach and Tournoux, 1988) are woefully inadequate as an anatomic reference but did encourage the use and further development of spatial normalization schemes to reduce size and shape variability of brains. Other atlases such as (Ono et al., 1990) and (Duvernoy, 1991, Schleicher et al., 2005) provide far more complete and detailed delineations but either do not provide any principles or algorithms for spatial normalization or Cartesian coordinate systems. And finally, the question of how a single brain based atlas can represent a population is rarely argued anymore.

Describe Where you are

What then can facilitate this need to compare and contrast brain image data? We believe this requires more complete descriptions. Just as when we give directions to our house we might say, “… the address is 123 Main street, it’s the third house on the left after the intersection, it’s a white house with a porch, two stories tall…”, etc. If using certain Global Positioning Systems (GPS) we might specify the latitude and longitude, in coordinates. We provide several ways to locate and identify the house. The different descriptions, collectively, make it easy to find and often compare it to others that may be geographically close. In brain mapping, this is difficult because often the only reference we have is an accompanying macroscopic structural MRI. So we might say the site of activation is in Brodmann area 46 or Talairach and Tournoux coordinate (x,y,z) or in the pars opercularis of the inferior frontal gyrus. We might say the activation occurred in the sulcus near the anterior temporal branch of the posterior cerebral artery or was concentrated in layer 4 of the cortex. None individually is very precise but in combination provide increasing degrees of specificity. Collectively, these descriptions make it easier to identify and compare the region where the activation occurred.

Integrate the data

What makes this notion of utilizing multifactorial descriptors intriguing, is the emergence of ever more novel classification schemes to help identify regions, circuits and systems of the brain. No longer are we confined to anatomic divisions defined by cell type or packing density observed using classical histological techniques. Complementing the cytoarchitectural descriptions of (Brodmann, 1909), (von Economo and Koskinas, 1925), (Duvernoy, 1991), (Ono et al., 1990), and (Mai J, 1997) are atlases of chemoarchitecture (Zilles et al., 2002), neurotransmitters (Schleicher et al., 2005), gene expression patterns (Mansour A, 1995), angioarchitecture (Nowinski et al., 2005), and white matter connections (Mori et al., 2005), among other things.

Unfortunately, most of these newer classification schemes are still works in progress. They may not cover the entire brain. They may still be based on one or few brains. They may not be compatible with methods to place them in correspondence with experimental data or they may have limited resolution, etc.

However, when combined, they can overcome their individual limitations. Although not yet available, efforts to computationally integrate these atlases are in progress ((Nowinski et al., 1997); (Maldjian et al., 2003); (Carmack et al., 2004); (Nowinski, 2004). Ontologies that provide relationship definitions between classification schemes are being created (Martin et al., 2003) and applications of historically important maps to interrogate modern MRI are now possible (see Figure 1). For these studies, elaborate surface-based and volumetric warping approaches have been developed to enable the transfer of classical maps into digital coordinates, and between digital atlases and individual datasets. Once available, the integration of these atlases will provide a far more specific and descriptive notion of what is where. After all, the separation of these classification schemes was borne out of necessity to collect the data, not because any one is more important than any other. Classification schemes each have a role to play as all are important. However, they should be integrated.

Figure 1
Elastic Normalization of the Brodmann and Schmahmann Atlases to Multisubject fMRI Data

Group Statistics

Can an atlas based on any one single individual be representative? The Devlin and Poldrack paper, makes a clear and convincing argument that this is inappropriate. Numerous other papers (Van Essen, 2005); (Toga et al., 2006)) have come to the same conclusion and many population based atlases have emerged in response to this need (Mazziotta et al., 2001). But the same logic can be carried to the next level by creating many different population based atlases, each specific to the group demographics, disease, age or other characteristics of the subjects being studied. These provide not only population statistics within the map but arguably better represent the morphological signature of that particular cohort. What must be included in all analyses are confidence statistics on where the activity takes place. Whether this entails a statistic on probability, percentile or other metric may depend on the experimental design and other factors.

Adoption of a single normal atlas, even a probabilistic version, for all subject studies provides for the nominal capability for easier comparisons but in doing so fails to adequately measure the nuances within each group.

Taking this discussion into consideration, it seems that it might be prudent to avoid dependency on a single modality, single group representation for every study as is argued in the Devlin and Poldrack paper. Any given imaging experiment will be better served by mapping to a population based atlas that closely resembles the cohort under study. We suggest that population atlases for groups such as Alzheimer’s (Thompson et al., 2000); (Mega et al., 2005), schizophrenia (Narr et al., 2001); (Cannon et al., 2006)), pediatric populations (Wilke et al., 2003); (Gogtay et al., 2004); (Evans and Group, 2006)), autism (Joshi et al., 2004), even decades of life (Mazziotta et al., 2001) should be utilized, as appropriate, for the subject group.

Standardization is premature

Another of the points made in the Devlin and Poldrack paper is that a practical solution must be available in order to encourage usage. The MNI atlas was identified because it is in digital format, represents a population rather than an individual, and is distributed as part of common image analysis packages that provide algorithms to align data to it. Although it is available, adoption as THE solution could have the unwanted consequence of reducing enthusiasm for development of more specific atlases. It also may further encourage adoption of yet another sanctioned solution to those with limited experience and understanding of the (in)appropriateness of a particular atlas for interrogation of functional anatomy. Perhaps the paper title; ‘In praise of Tedious Anatomy’ could be broadened to say ‘In Praise of Tedious Analysis’.

There are also ongoing developments in computational anatomy to create better canonical image templates to represent subpopulations, while retaining well-resolved anatomical features that are vital to assist automated algorithms for aligning data to them. These atlases may even have a time-varying component (Janke et al., 2001), allowing subjects of different ages to be brought into the atlas using an age-appropriate transformation. Rather than average images together voxel-by-voxel to produce a blurred template, as was done in many ‘first generation’ statistical atlases, many groups are developing practical methods to create well-resolved canonical atlas images that represent the statistical mean anatomy for a population, using deformation averaging (Collins et al., 1995); (Thompson et al., 2000); (Kochunov P, 2002); (Twining et al., 2005)), Lie group methods on deformation tensors (Woods, 2003); (Lepore et al., 2006)), or geodesics on groups of diffeomorphic flows (Joshi et al., 2004); (Miller et al., 2005); (Lorenzen et al., 2006)). These approaches are complex, but are advantageous as they are close (in a strictly defined mathematical sense) to the brains being normalized to them and are likely to improve spatial accuracy and reduce sources of bias when comparing datasets in a canonical coordinate system.

The utility of the atlas is only as good as the method used to map the data into it. Specifying the ‘where’ of functional activity requires a full and comprehensive description of how that was determined. What were the mapping algorithms that performed the registration or more likely the deformation to make the subject and the atlas spatially congruent? There are now many different approaches, with various limitations and assumptions, for mapping to an atlas. Some are based on volumetric registration, in which 3D transformations deform the full brain volume into correspondence with the atlas. These mappings were traditionally linear or piecewise linear (Talairach and Tournoux, 1988), but it is now more common to use polynomial, elastic or fluid transformations with thousands or millions of degrees of freedom to better adapt the individual data to the atlas (e.g., (Ashburner et al., 2000)). The principles guiding these registration approaches also continue to advance (e.g., (Chiang et al., 2006); (Leow et al., 2007)). Other approaches directly model the cortex, inducing a spherical or planar parametric reference grid onto the cortical surface, providing a mathematical framework to perform higher-order alignments of structure or function within the cortical sheet (Thompson et al., 2004); (Rasser et al., 2005)).

A likely innovation, over the coming years, is that surface-based and volumetric registration approaches may be routinely integrated when transferring data into an atlas coordinate space (Joshi et al., 2007). Despite their added complexity, surface-based approaches have some major advantages for integrating and localizing cortically-derived brain maps (e.g., (Fischl et al., 1998); (Zeineh et al., 2003); (Van Essen, 2005); (Goebel et al., 2006)) in that they respect the topology of the cortical sheet and allow improved alignment of functional, structural and architectonic maps across subjects and even across species. However, most brain mapping studies now use volumetric registration without explicit cortical modeling and without using cortical features as constraints, producing 3D statistical parametric maps that are difficult to cross-reference or reconcile with any representation of the cortex. Hybrid approaches using hierarchical deformations (e.g. (Thompson and Toga, 1996); (Joshi et al., 2007)) allow the full image volumes to be normalized along with the cortical surfaces and their internal sulcal landmarks, yielding spatially consistent statistical results in both the 3D volume and the cortical surface parameter space. These efforts may take some time to become routinely accepted, but the benefits of linking surface and volumetric data will make this integration highly probable. For these reasons, it is unlikely that any currently available standardized atlas will offer the generality and flexibility required to accommodate these future analytic developments.

Full Provenance

Methods sections of published papers are full of program descriptions but not necessarily the specific algorithms used, their parameters, or the preprocessing steps undertaken before the programs are executed. We need a way to capture the full history of processing, in all its detail, to truly make data comparable. There are now several efforts to include this information in databases (Toga, 2002). There are even programs that retain descriptors of the algorithms chosen as part of the results (Rex et al., 2003). These must be included as part of any atlasing approach. If this information is retained as a processing pipeline that is efficient to calculate, statistics from previously published studies can be recomputed after transferring data to new coordinate systems, facilitating cross-study comparisons and meta-analysis. Storage of information on data provenance is vital to ensure that the findings of old studies can be leveraged in the future. Future analytic developments in data normalization and atlasing can then be brought to bear on previously published data, even if the findings were originally reported in coordinate systems that are obsolete or incompatible with current or future ones. With data provenance, statistical maps previously reported in one coordinate system can be transferred to another, without requiring us to be dependent on any current atlas, now or in the future.

Knowledge of the anatomy may be tedious to learn but it is clearly needed and no less important than all the elements of the experiment and analysis, the imaging protocol, the mathematics, the reference(s) and the visualization.

Acknowledgments

This work was supported by the National Institutes of Health through the NIH Roadmap for Medical Research, grant U54 RR021813 entitled Center for Computational Biology (CCB). Information on the National Centers for Biomedical Computing can be obtained from <http://nihroadmap.nih.gov/bioinformatics>. Additional support was provided by the NIH research grants R01 LM005639, P01 EB001955, and P50 AG016570, the NIH/NCRR resource grant P41 RR013642, and the Biomedical Informatics Research Network (BIRN, http://www.nbirn.net).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Literature Cited

  • Ashburner J, Andersson JL, Friston KJ. Image registration using a symmetric prior - in three dimensions. Human Brain Mapping. 2000;9:212–225. [PubMed]
  • Brodmann K. Vergleichende lokalistionslehre der grosshirnrinde in ihren prinzipien dargestellt auf grund des zellenbause. In: Barth L, editor. Some Papers on the Cerebral Cortex. Thomas; Springfields, IL: 1909. pp. 201–230.
  • Cannon TD, Thompson PM, van Erp TG, Huttunen M, Lonnqvist J, Kaprio J, Toga AW. Mapping heritability and molecular genetic associations with cortical features using probabilistic brain atlases: methods and applications to schizophrenia. Neuroinformatics. 2006;4:5–19. [PubMed]
  • Carmack PS, Spence J, Gunst RF, Schucany WR, Woodward WA, Haley RW. Improved agreement between Talairach and MNI coordinate spaces in deep brain regions. NeuroImage. 2004;22:367–371. [PubMed]
  • Chiang MC, Dutton RA, Hayashi KM, Toga AW, Lopez OL, Aizenstein HJ, Becker JT, Thompson PM. 3D pattern of brain atrophy in HIV/AIDS mapped using tensor-based morphometry. Proceedings of the Annual Meeting of the American Academy of Neurology, pp.2006.
  • Collins DL, Holmes CJTMP, Evans AC. Automatic 3-D model based neuroanatomical segmentation. Human Brain Mapping. 1995;3:190–208.
  • Duvernoy HM. The human brain. Springer-Verlag; New York: 1991.
  • Evans AC, Group BDC. The NIH MRI study of normal brain development. NeuroImage. 2006;30:184–202. [PubMed]
  • Fischl B, Sereno MI, Tootell RBH, Dale AM. High-resolution inter-subject averaginga coordinate system for the cortical surface. Human Brain Mapping. 1998;8:272–284. [PubMed]
  • Goebel R, Esposito F, Formisano E. Analysis of functional image analysis contest (FIAC) data with brainvoyager QX: From single-subject to cortically aligned group general linear model analysisself-organizing group independent component analysis. Hum Brain Mapp. 2006;27:392–401. [PubMed]
  • Gogtay N, Giedd JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis AC, Nugent TF, 3rd, Herman DH, Clasen LS, Toga AW, Rapoport JL, Thompson PM. Dynamic mapping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences of the United States of America. 2004;101:8174–8179. [PMC free article] [PubMed]
  • Holmes CJHR, Collins L, Woods R, Toga AW, Evans AC. Enhancement of MR images using registration for signal averaging. J Comput Assist Tomogr. 1998;22:324–333. [PubMed]
  • Janke AL, Zubicaray GI, Rose SE, Griffin M, Chalk JB, Galloway GJ. 4D deformation modeling of cortical disease progression in Alzheimer's dementia. Magn Reson Med. 2001;46:664–666. [PubMed]
  • Joshi AA, Shattuck DW, Thompson PM, Leahy RM. Simultaneous surface and volumetric brain registration using harmonic mappings. IEEE Trans Med Imag. 2007 In press. [PubMed]
  • Joshi S, Davis B, Jomier M, Gerigb G. Unbiased diffeomorphic atlas construction for computational anatomy. NeuroImage. 2004;23:S151–S160. [PubMed]
  • Kochunov PLJ, Thompson P, Toga AW, Brewer P, Hardies J, Fox P. An optimized individual target brain in the talairach coordinate system. NeuroImage. 2002;17:922–927. [PubMed]
  • Leow AD, Yanovsky I, Chiang MC, Lee AD, Lu A, Klunder AD, Becker JT, Davis SW, Toga AW, Thompson PM. Statistical properties of Jacobian maps and inverse-consistent deformations in non-linear image registration. IEEE Trans Med Imag. 2007 In Press. [PubMed]
  • Lepore N, Brun CA, Chiang MC, Chou YY, Dutton RA, Hayashi KM, Becker JT, Aizenstein HJ, Lopez OL, Toga AW, Thompson PM. Multivariate statistics of the Jacobian matricies in tensor-based morphometry and their application to HIV/AIDS. Proceedings of the 9th IEEE Conference on Medical Image Computer and Computer Assisted Intervention (MICCAI2006); Copenhagen, Denmark. 2006. [PubMed]
  • Lorenzen P, Prastawa M, Davis B, Gerig G, Bullitt E, Joshi S. Multimodal image set registration and atlas formation. Med Image Anal. 2006;10:440–451. [PMC free article] [PubMed]
  • Mai JAJ, Paxinos G. Atlas of the human brain. Academic; New York: 1997.
  • Makris N, Hodge SM, Haselgrove C, Kennedy DN, Dale A, Fischl B, Rosen BR, Harris G, Caviness VS, Schmanhmann JD. Human cerebellum: surface-assisted cortical parcellation and volumetry with magnetic resonance imaging 2003 [PubMed]
  • Maldjian JA, Laurienti PJ, Kraft RA, Burdette JH. An automated method for neuroanatomical and cytoarchitectonic atlas-based interrogation of fMRI data sets. NeuroImage. 2003;19:1233–1239. [PubMed]
  • Mansour AFC, Burke S, Akil H, Watson SJ. Immunohistochemical Localization of the Cloned Mu Opioid Receptor in the Rat CNS. J Chem Neuroanat. 1995;8:283–305. [PubMed]
  • Martin RF, Rickard K, Mejino JL, Agoncillo AV, Brinkley JF, Rosse C. Structural informatics group, the evolving neuroanatomical component of the foundational model of anatomy. AMIA Annu Symp Proc. 2003;2003:927. [PMC free article] [PubMed]
  • Mazziotta J, Toga A, Evans A, Fox P, Lancaster J, Zilles K, Woods R, Paus T, Simpson G, Pike B, Holmes C, Collins L, Thompson P, MacDonald D, Iacoboni M, Schormann T, Amunts K, Palomero-Gallagher N, Geyer S, Parsons L, Narr K, Kabani N, Le Goualher G, Boomsma D, Cannon T, Kawashima R, Mazoyer B. A probabilistic atlas and reference system for the human brain: International Consortium for Brain Mapping (ICBM. Philos Trans R Soc Lond B Biol Sci. 2001;356:1293–1322. [PMC free article] [PubMed]
  • Mega MS, Dinov ID, Mazziotta JC, Manese M, Thompson PM, Lindshield C, Moussai J, Tran N, Olsen K, Zoumalan CI, Woods RP, Toga AW. Automated brain tissue assessment in the elderly and demented population: construction and validation of a sub-volume probabilistic brain atlas. Neuroimage. 2005;26:1009–1018. [PubMed]
  • Miller MI, Beg MF, Ceritoglu C, Stark C. Increasing the power of functional maps of the medial temporal lobe by using large deformation diffeomorphic metric mapping. Proc Natl Acad Sci U S A. 2005;102:9685–9690. [PMC free article] [PubMed]
  • Mori S, Wakana S, Nagae-Poetscher LM, Van Zijl PC. MRI atlas of human white matter. Elsevier; Amsterdam, The Netherlands: 2005.
  • Narr KL, Thompson PM, Sharma T, Moussai J, Blanton R, Anvar B, Edris A, Krupp R, Rayman J, Khaledy M, Toga AW. Three-dimensional mapping of temporo-limbic regions and the lateral ventricles in schizophrenia: gender effects. Biol Psychiatry. 2001;50:84–97. [PubMed]
  • Nowinski WL. Co-registration of the Schaltenbrand-Wahren microseries with the probabilistic functional atlas. Stereotact Funct Neurosurg. 2004;82:142–146. [PubMed]
  • Nowinski WL, Fang A, Nguyen BT, Raphel JK, Jagannathan L, Raghavan R, Bryan RN, Miller GA. Multiple brain atlas database and atlas-based neuroimaging system. Comput Aided Surg. 1997;2:42–66. [PubMed]
  • Nowinski WL, Thirunavuukarasuu A, Volkau I, Baimuratov R, Hu Q, Aziz A, Huang S. Informatics in radiology (infoRAD): three-dimensional atlas of te brain anatomy and vasculature. Radiographics. 2005;25:263–271. [PubMed]
  • Ono M, Kubik S, Abernathey CD. Atlas of the cerebral sulci. Thieme Verlag; Stuttgart New York: 1990.
  • Rasser P, Johnston P, Lagopoulos J, Ward P, Schall U, Thienel R, Bender S, Toga A, Thompson P. Functional MRI BOLD response to Tower of London performance of first-episode schizophrenia patients using cortical pattern matching. NeuroImage. 2005;26:941–951. [PubMed]
  • Rex D, Ma J, Toga A. The LONI pipeline processing environment. NeuroImage. 2003;19:1033–1048. [PubMed]
  • Schleicher A, Palomero-Gallagher N, Morosan P, Eickhoff S, Kowalski T, de Vos K, Amunts K, Zilles K. Quantitative architectonic analysis: A new approach to cortical mapping. Anat Embryol. 2005;210:373–386. [PubMed]
  • Talairach J, Tournoux P. Co-planar stereotaxic atlas of the human brain: 3-dimensional proportional system - an approach to cerebral imaging. Thieme Medical Publishers; New York: 1988.
  • Thompson P, Hayashi K, Sowell E, Gogtay N, Giedd J, Rapoport J, de Zubicaray G, Janke A, Rose S, Semple J, Doddrell D, Wang Y, van Erp T, Cannon T, Toga A. Mapping cortical change in Alzheimer's disease, brain development, and schizophrenia. NeuroImage. 2004;23:S2–18. [PubMed]
  • Thompson P, Toga A. A Surface-based technique for warping 3-dimensional brain. IEEE Transactions on Medical Imaging. 1996;15:1–16. [PubMed]
  • Thompson P, Woods R, Mega M, Toga A. Mathematical/computational challenges in creating deformable and probabilistic atlases of the human brain. Human Brain Mapping. 2000;9:81–92. [PubMed]
  • Toga A. Neuroimaging Databases: The Good, the Bad and the Ugly. Nature Reviews Neuroscience. 2002;3:302–309. [PubMed]
  • Toga AW, Thompson PM, Mori S, Amunts K, Zilles K. Towards multimodal atlases of the human brain. Nature Reviews Neuroscience. 2006;7:952–966. [PMC free article] [PubMed]
  • Twining CJ, Cootes T, Marsland S, Petrovic V, Schestowitz R, Taylor CJ. A unified information-theoretic approach to groupwise non-rigid registration and model building. Proceedings of the Information Processing in Medical Imaging (IPMI); 2005; Glenwood Springs, CO. 2005. [PubMed]
  • Van Essen DC. A population-average, landmark- and surface-based (PALS) atlas of human cerebral cortex. NeuroImage. 2005;28:635–662. [PubMed]
  • von Economo C, Koskinas GN. Die cytoarchitektonik der hirnrinde des erwachsenen menschen. Springer-Verlag; Berline: 1925.
  • Wilke M, Schmithorst VJ, Holland SK. Normative pediatric brain data for spatial normalization and segmentation differs from standard adult data. Magn Reson Med. 2003;50:749–757. [PubMed]
  • Woods RP. Characterizing volume and surface deformations in an atlas framework: theory, applications, and implementation. NeuroImage. 2003;18:768–788. [PubMed]
  • Zeineh MM, Engel SA, Thompson PM, Bookheimer SY. Dynamics of the hippocampus during encoding and retrieval of face-name pairs. Science. 2003;299:577–580. [PubMed]
  • Zilles K, Palomero-Gallagher N, Grefkes C, Scheperjans F, Boy C, Amunts K, Schleicher A. Architectonics of the human cerebral cortex and transmitter receptor fingerprints: Reconciling functional neuroanatomy and neurochemistry. Europe Neuropsycholpharmacol. 2002;12:587–599. [PubMed]
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • PubMed
    PubMed
    PubMed citations for these articles