Anatomical Parcellation of Cortical Language Sites
Richard F. Martin, PhD, James F. Brinkley, MD, PhD, Xenia Hertzenberg, Andrew Poliakov, PhD, David Corina, PhD and George Ojemann, MD
Structural Informatics Group, Departments of Biological Structure and Medical Education and Biomedical Informatics,
Computer Science and Engineering, Psychology and Neurosurgery
University of Washington, Seattle WA 98195
Supported by Human Brain Project grant DC02310 and NIH grant LM06822
Query Output: Visualizations on
Cortical Parcellation Maps
Foundational Model of Neuroanatomy
•Used for queries about regions of the brain
that are affected for certain language errors.
• Map regions of the brain that appear in the
result.
1)
• Two types of visualizations:
1) Localization of result by brain parcel
2) Rate of occurrence by brain parcel
•Provide further filtering capabilities.
Figure 1: Photograph of neurosurgical exposure of
lateral surface of left hemisphere of cerebral cortex
showing CSM sites (numbered labels).
Introduction:
Figure 2: Screenshot of SKANDHA Brain Mapper
interface used to plot location of CSM sites on 3D
models of individual patients’ brains.
Figure 5: FMA Protégé authoring environment
showing symbolic representation of subdivisions of
cerebral cortical gyri.
• Works for certain classes of user queries.
2)
Neural Component of Foundational Model of Anatomy (FMA):
An ontology of neuroanatomy to promote data integration and sharing
Anatomical labeling of cerebral cortical stimulation (CSM) sites is
necessary for intelligent computer querying of a rich and unique
experimental database examining neural substrates underlying human
language production. We have developed a parcellation scheme for
the lateral surface of the human cerebral cortex. We then compared
results generated utilizing this approach to those generated using an
alternative method implemented in the Talairach Daemon.
 A consistent naming system correlated with existing terminologies such as
NeuroNames and Terminologia Anatomica
 A logical and consistent framework based on inheritance of structural attributes
 Scales from the molecular to macroscopic levels
 Accommodates diverse types of attributes of normal structures
 Integrated in UW Foundational Model of Anatomy of the whole body (FMA)
 Implemented in the Protégé frame-based knowledge acquisition system
Mapping brain functional anatomy using electrical stimulation of
cerebral cortex in awake neurosurgical patients provides a unique
window into the neurological substrate underlying human language.
The complex and varied nature of data sets collected during these
procedures has necessitated development of sophisticated
experiment management systems (EMS)
http://sig.biostr.washington.edu/projects/brain/ for storage and
analysis of demographic, behavioral, neurophysiological, functional
Magnetic Resonance and neuroanatomical data. The precise
anatomical location of each CSM site has proven to be the single
data point which best links all other data types. Whereas graphical
representations of CSM sites plotted on 3-D MR-based
reconstructions are needed for visualization and quantitative
analysis, efficient retrieval necessitates assignment of text-based
anatomical labels to the CSM sites.
------------------------------------------------------------------------------------------------------
Figure 3: Cortical Parcellation System – subdividing the
lateral surface of the human cerebral cortex. CSM sites
are each assigned to corresponding anatomical parcel.
To this end, we have developed a scheme to parcellate the lateral
surface of the human cerebral cortex into thirty-seven
subdivisions, labeled using Foundational Model of Anatomy
(FMA) expansion of NeuroNames terminology1 (Fig. 5) We call
this scheme the Cortical Parcellation System (CPS). Each CSM
site is assigned to the appropriate parcel by an expert, using
anatomical landmarks visible on the cortical surface.
Figure 6: Example of use of CPS to
illustrate relative locations of cortical sites
associated with language tasks requiring
subjects to provide either a verb or a noun
in response to a visual symbol.
Once parcellation assignments are made for each CSM site, it is possible to generate queries of the EMS
database across patients based primarily on anatomical parcels of interest, as presented in a companion
poster by Z. Tang et al. (Poster #1-055)
To test the efficacy of this parcellation method we compared, on thirty-one patients, labels generated using
our method versus those generated using the Talairach Daemon (TD) http://ric.uthscsa.edu/projects/,
access to which we have incorporated in our EMS. Out of a total of 511 CSM sites compared, only 306 sites
(60%) were assigned to the same cortical gyrus by both the TD and the CPS parcellation scheme.
Agreement was as high as 92% in one patient, and as low as 19% in another. We attribute these differences
to the considerable individual variation of cortical surface landmarks which demarcate gyral boundaries.
Because TD relies on relatively distant deep brain landmarks (AC–PC plane) to coregister brains, we
submit that the TD technique is disadvantaged with regard to determination of cortical gyral boundaries
when compared to our parcellation technique, which relies on visual inspection of cortical surface
landmarks on each individual brain. In future work, we will extend our analysis to determine what factors
affect the degree of agreement between these and other anatomical assignment techniques.
Figure 7: Example of use of CPS to graphically display results of
XBrain query submitted to the EMS database. This query requested
the anatomical location and incidence of Type 2 (Phonological
paraphasia) language errors elicited by bipolar stimulation of cerebral
cortex across the entire patient database.
Conclusions:
The Cortical Parcellation System presented here provides a useful means of
assigning anatomical location information to cortical stimulation mapping
site data sets obtained from human patients. The CPS allows for more
accurate anatomical localization of CSM sites due to the system’s emphasis
on localization based on cerebral cortical landmarks, rather than deep brain
structures. Accurate anatomical localization of CSM sites facilitates multipatient analysis of cortical physiological data and testing of hypotheses
related to understanding language production in humans.
[1] Martin, R. F., Mejino, J. L. V., Bowden, D. M., Brinkley, J. F. and Rosse, C. (2001) Foundational
Model of Neuroanatomy: Implications for the Human Brain Project. In Proceedings, AMIA Annual
Fall Symposium, pages 438-442, Washington, DC.
For more information, e-mail: < fma@sig.biostr.washington.edu >