Wavelet-Based Independent Component Analysis For Statistical Shape Modeling

Over the past decade, active shape models have gained increased popularity in medical image analysis. However, despite its widespread, it is now widely accepted that classical shape models using principle component analysis (PCA) is not able to faithfully model the wide range of variations that anatomical structures can undergo. In this paper, we present a new statistical shape model using wavelet transform and independent component analysis (ICA). In an attempt to benefit from the sparsification and approximation power of wavelets, we investigate constructing an ICA-based shape model in a compressed wavelet domain. In order to assess the efficiency of the proposed shape model; experiments were conducted using contours of human vertebrae from X-ray images.