HEp-2 cells staining patterns classification via wavelet scattering network and random forest

The classification of HEp- 2 cells staining patterns is a particularly useful task in identifying various autoimmune diseases in medical applications. Conventionally, this task is carried out by specialists' observing under a fluorescence microscope, which depends too heavily on the experience and expertise of the operators. Due to the complexity of these stain patterns, their identification process is often subjective and unreliable [2]. Manual identification of these patterns from a large collection of cell images is very laborious, and often suffers from intrinsic limitations related to visual evaluation performed by human [8]. To overcome these limitations, over the past few years, the automatic classification of HEp- 2 cells staining patterns (as shown in Fig. 1) has attracted increasing attention in computer vision research and various Computer-Aided Diagno-sis(CAD) systems have been designed with image analysis techniques to reduce the labor and time required by the analysis [9], [7]. Although tremendous progress has been made towards improving the identification accuracy of the cell stain patterns, the performances of the state-of-the-art systems are still far from its use by medical practitioners. Thus, automatic classification of the staining patterns has been increasingly demanded.