Parametric representation of cortical surface folding based on polynomials

The development of folding descriptors as an effective approach for describing geometrical complexity and variation of the human cerebral cortex has been of great interests. This paper presents a parametric representation of cortical surface patches using polynomials, that is, the primitive cortical patch is compactly and effectively described by four parametric coefficients. By this parametric representation, the patterns of cortical patches can be classified by either model-driven approach or data-driven clustering approach. In the model-driven approach, any patch of the cortical surface is classified into one of eight primitive shape patterns including peak, pit, ridge, valley, saddle ridge, saddle valley, flat and inflection, corresponding to eight sub-spaces of the four parameters. The major advantage of this polynomial representation of cortical folding pattern is its compactness and effectiveness, while being rich in shape information. We have applied this parametric representation for segmentation of cortical surface and promising results are obtained.