A multiscale decomposition approach to gel image interpretation

Abstract

Two-dimensional (2-D) polyacrylamide gel electrophoresis (PAGE) images are widely used in the area of Proteomics. In this paper, a region-based multiscale decomposition approach to gel image interpretation is presented. Instead of segmentation of each individual protein spot, the gel image is decomposed into small local regions with a quadtree structure. 2-D Gaussian functions are used to model local image features. Those local regional features will be used for further statistical analysis to extract discriminant features that distinguish different classes of gels. After a description of the method, experimental results are presented to show its potential in representing the protein related pixel intensities of the gel images. © Springer-Verlag 2003.