|Home | About | Journals | Submit | Contact Us | Français|
The geometry of the human cerebral cortex is very complex and variable across individuals. An essential characteristic of cortical geometry is its folding. Though the mechanisms of cortical folding are still largely unknown , it was reported that cortical folding pattern is a good predictor of brain function . Hence, development of cortical folding pattern descriptors could potentially contribute to automated segmentation and recognition of brain anatomies, as well as the mapping of brain structure and its function. There already exist a couple of cortical folding descriptors in the literature, e.g., curvature , gyrification index , and spherical wavelet . These folding descriptors have their own advantages and are already applied in many research and clinical studies [6, 7].
This paper presents a parametric folding pattern descriptor for the cortex at the meso-scale of primitive surface patch. The key idea of this descriptor is that we intend to encode geometric shape pattern information of surface patch by the parametric Bezier surface representation. Figure 1 provides an overview of the proposed method. Main steps are followed:
An important contribution of the BCP image representation is that the analysis of surface patch folding pattern is converted into the problem of image pattern classification, which has been extensively studied in the computer vision and pattern recognition community and there are many readily usable algorithms and tools for image pattern analysis [9, 10]. In our experiments we use Locality Preserving Projections (LPP) method. The essence of this approach is to map the original BCP images space to a subspace expanded by the eigenfunctions of the Laplace Beltrami operator on the BCP manifold. Using above pipeline we make the classification on synthesized dataset, random sulci and gyri patches and different sulci regions extracted from human cortical surfaces. The correctness of classification is very satisfied. Figure 4 and figure 5 show the examples of our manually extracted patches.
In summary, the major contribution of this paper is the development of BCP image as a folding shape pattern descriptor. In this method, the variable shapes of surface patches are compactly and effectively encoded by the regular grids of BCP images. The classification of surface folding pattern problem is converted to a regular image pattern classification problem using the LPP projection method. Though this work and its results are preliminary, we believe this methodology has great potential to be extended for more complex cortical folding analysis in the future. This effective folding pattern representation methodology could be potentially used for many applications in cortical surface analysis such as automated parcellation and recognition of cortical surface, computational neuroanatomy, and clinical studies of brain diseases with abnormal cortical folding patterns.