Staining Pattern Recognition Of HEp-2 Cell Images Using Supervised Canonical Correlation Analysis

The objective of this paper is to develop a methodology for automatic recognition of antinuclear antibodies by human epithelial type 2 (HEp-2) cell indirect immunofluorescence (IIF) image analysis to diagnose connective tissue disease. It integrates judiciously the merits of a new supervised canonical correlation analysis, termed as CuRSaR, and the theory of rough hypercuboid approach. The proposed method efficiently combines the textural information of HEp-2 cells, derived from local binary patterns and its variants, by canonical correlation analysis, while significant and relevant features for HEp-2 cell classification are extracted using the rough hypercuboid approach. Finally, the support vector machine is used to recognize one of the known staining patterns in IIF images. The effectiveness of the proposed method, along with a comparison with related feature extraction and multimodal data integration methods, is demonstrated on a set of HEp-2 cell images.