Multi-Scale Analysis of Brain Surface Data

Abstract

The human brain is characterized by complex convolution patterns. Analyzing the variability of these patterns among human subjects can reveal information for the detection of diseases that affect the human brain. This chapter presents a novel method to visualize the brain surface and its folding pattern at different scales. The analysis steps involved are the transformation of the cortical surface from high resolution magnetic resonance tomography images (MRI) to an initial representation as a triangulated mesh and finally to a representation as a series of spherical harmonic basis functions. The spherical harmonic parameterization of the surface is translation, rotation and scaling invariant. The parametric representation gives a multidimensional coefficient vector for each cortical surface. The technique allows easier recognition of convolutional patterns. The method is a first step toward a statistical multi-scale analysis of the brain surface.