The cerebral cortex makes up the largest part of the human brain. Each cortical hemisphere has the topology of a folded, 2-D sheet. Nevertheless, most approaches to analyzing and displaying human brain imaging data have relied upon volume-based (3-D) approaches. There are many advantages to visualizing and analyzing data using a system based on reconstructing the cortical surface. The reason this technique has not been more widely used is the difficulty of automating the reconstruction process. A software system that can do this is described. A high-resolution MRI data set is first optimized for gray white-matter contrast. The skull is automatically stripped off, and the data set is anisotropically filtered. The cortical white matter of each hemisphere is then filled by 3-D region growing and tessellated to reconstruct the gray- white surface. This surface can be used to sample functional data (fMRI or PET) or accurately measure cortical thickness. The surface can then be unfolded to bring the depths of sulci into view; or it can be cut to allow complete flattening so that the entire cortical surface of ahemisphere can be viewed at once. Finally, images of this surface are rendered for display.
Specific aims for Phase 1 are: (1) to verify and improve the accuracy of the existing reconstruction and unfolding method, (2) provide better support for open standards for input and output image data formats, (3) to port the display code to the hardware- and window-system-independent OpenGL API, and (4) to build a user friendly point-and-click inter-face and scripting language based on Tk/Tcl. A primary use of an automatic cortical surface reconstruction method is to analyze and visualize functional imaging data from fMRI and PET, and to provide a framework for merging data from different imaging modalities.A cortical surface makes possible quantitative measures local cortical thickness (useful for assessing cortical pathology). Visualizing the gray- white matter surface can help with surgical planning. Finally, the cortical surface can also provide constraints for MEG/EEG source imaging.
There are no existing non-invasive methods for automatically reconstructing the human cortical surface. The proposed product would be of interest to most research and many clinical groups studying the human brain using MRI and PET. There are over 8000 MRI scanners in existence. They support a much larger number of users. This software package would initially be targeted to researchers in human brain imaging.
