Hierarchical classification framework for HEp-2 cell images

Currently, indirect immune fluorescence imaging of human epithelial type 2 (HEp-2) cell image is an effective evidence to diagnose autoimmune diseases. In this work, a novel hierarchical classification model is developed for cell images. To be more specific, in the first step, the six-class task is constructed as a two-class task, where five categories are merged into one category, except the one that is the most difficult to distinguish. After that, the second step is to distinguish the combined five categories. During this process, Codebook less model (CLM) is used to extract the characteristics of the images. Feature mapping is used to effectively narrow the gap between training sets and test sets. Hierarchical classification framework is evaluated systematically on the IEEE International Conference on Image Processing (ICIP) 2013 contest dataset. Experimental results demonstrate the effectiveness of the proposed method.