11th Annual OHBM Meeting

Abstract Number: 633
Submitted By: Aaron Kline
Last Modified: 11 Jan 05

Comparison of Human Cortical Surface Reconstructions from Magnetic Resonance Imaging Data
Aaron Kline, Deborah Smith, Monica K. Hurdal
Department of Mathematics, Florida State University, Tallahassee, U.S.A., 32306-4510

Objective: Increasingly, brain cartographic methods are being implemented in attempts to create cortical surface reconstructions that are useful and accurate. Aesthetically pleasing images of the human cortical surface have been rendered, but beyond visual appeal, these surfaces need to be characteristically accurate. Creating characteristically accurate surfaces is a nontrivial objective, for they must be topologically correct and accurately represent cerebral anatomy. A comparison of the characteristics of the surfaces created by packages such as FreeSurfer[1], BrainVISA[2], and others will aide in identifying strengths of the different software packages.

Methods:
We are evaluating cortical surface reconstruction methods contained in a number of software packages that are currently available to the neuroscience community. Several largely automated algorithms have been implemented in an attempt to create accurate human cortical surface reconstructions, three of which are FreeSurfer, BrainVISA and INCSurf[3]. We have created surfaces representing the grey and white matter using these different methods on magnetic resonance imaging (MRI) data from 11 different subjects. We are examining a number of different surface characteristics of the cerebral hemisphere, including surface area, sulcal length, and maximum volume bound of the cortex. The maximum volume bound represents the maximum volume of the cortex if the sulci were "filled" in.

Results & Discussion:
Preliminary attributes of the surfaces, such as surface area, vary considerably in the surfaces of single subjects across methods (see Table 1).

Conclusions:
Studying and quantifying different cortical surface reconstruction algorithms is an important issue for neuromorphometric analysis. It is critical to understand the biases, differences, and similarities in the resulting surfaces produced from different algorithms in this study. These validated surface reconstructions will aide in more localized and accurate identification of cerebral processing, opening the door for more insightful neurological studies.

References & Acknowledgements:
[1] Dale, A. M., Fischl, B., and Sereno, M. I., Cortical Surface-based Analysis: I. Segmentation and Surface Reconstruction. NeuroImage, vol. 9: 179-194, 1999. FreeSurfer available at http://surfer.nemr.mgh.harvard.edu/
[2] Mangin, J. F., Riviere, D., Cachia, A., Duchesnay, E., Cointepas, Y., Papadopoulos-Orfanos, D., Collins, D. L., Evans, A. C., Regis, J., Object-based Morphometry of the Cerebral Cortex. IEEE Trans. Medical Imaging, 23(8):968-982, Aug. 2004. BrainVISA available at http://brainvisa.info/
[3] Stern, J., Schaper, K., Rehm, K., Hurdal, M. K., Sumners, De Witt, and Rottenberg, D. A., Automatic Surface Extraction by Discrete, Topolgically-controlled, Region Growing. NeuroImage, vol. 13, pg. S44, 2001.

This work is supported in part by NSF grant DMS-0101329 and NIH grant P20 EB02013. We would like to thank Dr. David Rottenberg, Departments of Radiology and Neurology, University of Minnesota for providing the MRI data.

Table 1
S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11
FS pia 96785 65428 83639 65362 85052 74720 71457 62693 76606 81432 79119
FS wm 110921 85400 107206 83471 117507 96530 93154 79187 100255 105236 102964
BV pia 78606 73664 92494 76441 99246 86887 75083 65817 86714 83729 85676
BV wm 70389 65926 83576 64022 84358 78125 68644 54337 76935 73704 79568
INC pia 183095 153157 197869 84909 208727 158563 159311 147548 181649 187036 166152
FS = FreeSurfer, BV = BrainVISA, INC = INCSurf, wm = white matter, numbers in mm^2 for left hemisphere
NeuroImage, Volume 26, Supplement 1, Page S37, CD-Rom Abstract 633, 2005