Nonparametric Density Estimation and Discrimination from Images of Shapes
Abstract:Nonparametric density estimation is the basis for a new methodology for discrimination using shape data in the form of pixel images. Our work is driven by an application based on screening for neural tube defects from ultrasonography data that comprise binary pixel images of head shapes from human fetuses. We discuss the choice of smoothing parameters used for the density estimates, the variation that is inherent in our method and how our approach could be extended to take into account other discriminatory information. We compare our method based on density estimates with alternative approaches such as those based on Fourier descriptors.
Document Type: Original Article
Publication date: 1997-01-01