Characterization of neuroanatomic structures using 3D wavelet-based normal fields

This paper presents a new approach for the morphometrical characterization of brain structures using shape features calculated from normal fields. An orientable surface representation of the brain structure is derived allowing the estimation of the corresponding normal field without the need of a parameterized surface. In fact, the normal field is calculated from the application of the 3D wavelet transform. Rotation- and translation-invariant shape features extracted from the normal fields are defined and successful experimental results using data from neuroanatomic structures are presented.